├── src ├── nodes │ ├── newyear_node.py │ ├── __init__.py │ ├── hat_effect_node.py │ └── firecracker_effect_node.py └── __init__.py ├── model_configs └── mmpose │ ├── _base_ │ ├── filters │ │ ├── gaussian.py │ │ ├── one_euro.py │ │ ├── savizky_golay.py │ │ ├── smoothnet_t8_h36m.py │ │ ├── smoothnet_t16_h36m.py │ │ ├── smoothnet_t32_h36m.py │ │ ├── smoothnet_t64_h36m.py │ │ └── smoothnet_h36m.md │ ├── default_runtime.py │ └── datasets │ │ ├── deepfashion_lower.py │ │ ├── deepfashion_upper.py │ │ ├── deepfashion_full.py │ │ ├── zebra.py │ │ ├── aflw.py │ │ ├── cofw.py │ │ ├── jhmdb.py │ │ ├── atrw.py │ │ ├── mpi_inf_3dhp.py │ │ ├── campus.py │ │ ├── shelf.py │ │ ├── crowdpose.py │ │ ├── aic.py │ │ ├── ap10k.py │ │ ├── h36m.py │ │ ├── mpii.py │ │ ├── mhp.py │ │ ├── interhand2d.py │ │ ├── onehand10k.py │ │ ├── panoptic_body3d.py │ │ ├── panoptic_hand2d.py │ │ ├── freihand2d.py │ │ ├── posetrack18.py │ │ ├── ochuman.py │ │ ├── animalpose.py │ │ ├── macaque.py │ │ ├── rhd2d.py │ │ ├── horse10.py │ │ └── fly.py │ ├── vipnas_mbv3_coco_wholebody_256x192_dark.py │ └── hrnet_w32_animalpose_256x256.py ├── configs └── examples │ ├── pose_estimation │ ├── test_camera.py │ ├── pose_tracking.py │ ├── webcam_demo.md │ └── pose_estimation.py │ ├── new_year │ ├── README.md │ └── new_year.py │ ├── valentine_magic │ ├── README.md │ └── valentine_magic.py │ ├── bing_dwen_dwen │ ├── README.md │ ├── bing_dwen_dwen.py │ └── resource-info.json │ └── face_swap │ ├── README.md │ └── face_swap.py ├── .pre-commit-config.yaml ├── README.md ├── README_en.md ├── run.py └── .gitignore /src/nodes/newyear_node.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /src/__init__.py: -------------------------------------------------------------------------------- 1 | from . import nodes # noqa 2 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/gaussian.py: -------------------------------------------------------------------------------- 1 | filter_cfg = dict( 2 | type='GaussianFilter', 3 | window_size=11, 4 | sigma=4.0, 5 | ) 6 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/one_euro.py: -------------------------------------------------------------------------------- 1 | filter_cfg = dict( 2 | type='OneEuroFilter', 3 | min_cutoff=0.004, 4 | beta=0.7, 5 | ) 6 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/savizky_golay.py: -------------------------------------------------------------------------------- 1 | filter_cfg = dict( 2 | type='SavizkyGolayFilter', 3 | window_size=11, 4 | polyorder=2, 5 | ) 6 | -------------------------------------------------------------------------------- /src/nodes/__init__.py: -------------------------------------------------------------------------------- 1 | from .bingdwendwen_node import BingDwenDwenNode 2 | from .faceswap_node import FaceSwapNode 3 | from .firecracker_effect_node import FirecrackerEffectNode 4 | from .hat_effect_node import HatEffectNode 5 | from .valentine_magic_node import ValentineMagicNode 6 | 7 | __all__ = [ 8 | 'BingDwenDwenNode', 'FaceSwapNode', 'ValentineMagicNode', 'HatEffectNode', 9 | 'FirecrackerEffectNode' 10 | ] 11 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/smoothnet_t8_h36m.py: -------------------------------------------------------------------------------- 1 | # Config for SmoothNet filter trained on Human3.6M data with a window size of 2 | # 8. The model is trained using root-centered keypoint coordinates around the 3 | # pelvis (index:0), thus we set root_index=0 for the filter 4 | filter_cfg = dict( 5 | type='SmoothNetFilter', 6 | window_size=8, 7 | output_size=8, 8 | checkpoint='https://download.openmmlab.com/mmpose/plugin/smoothnet/' 9 | 'smoothnet_ws8_h36m.pth', 10 | hidden_size=512, 11 | res_hidden_size=256, 12 | num_blocks=3, 13 | root_index=0) 14 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/smoothnet_t16_h36m.py: -------------------------------------------------------------------------------- 1 | # Config for SmoothNet filter trained on Human3.6M data with a window size of 2 | # 16. The model is trained using root-centered keypoint coordinates around the 3 | # pelvis (index:0), thus we set root_index=0 for the filter 4 | filter_cfg = dict( 5 | type='SmoothNetFilter', 6 | window_size=16, 7 | output_size=16, 8 | checkpoint='https://download.openmmlab.com/mmpose/plugin/smoothnet/' 9 | 'smoothnet_ws16_h36m.pth', 10 | hidden_size=512, 11 | res_hidden_size=256, 12 | num_blocks=3, 13 | root_index=0) 14 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/smoothnet_t32_h36m.py: -------------------------------------------------------------------------------- 1 | # Config for SmoothNet filter trained on Human3.6M data with a window size of 2 | # 32. The model is trained using root-centered keypoint coordinates around the 3 | # pelvis (index:0), thus we set root_index=0 for the filter 4 | filter_cfg = dict( 5 | type='SmoothNetFilter', 6 | window_size=32, 7 | output_size=32, 8 | checkpoint='https://download.openmmlab.com/mmpose/plugin/smoothnet/' 9 | 'smoothnet_ws32_h36m.pth', 10 | hidden_size=512, 11 | res_hidden_size=256, 12 | num_blocks=3, 13 | root_index=0) 14 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/smoothnet_t64_h36m.py: -------------------------------------------------------------------------------- 1 | # Config for SmoothNet filter trained on Human3.6M data with a window size of 2 | # 64. The model is trained using root-centered keypoint coordinates around the 3 | # pelvis (index:0), thus we set root_index=0 for the filter 4 | filter_cfg = dict( 5 | type='SmoothNetFilter', 6 | window_size=64, 7 | output_size=64, 8 | checkpoint='https://download.openmmlab.com/mmpose/plugin/smoothnet/' 9 | 'smoothnet_ws64_h36m.pth', 10 | hidden_size=512, 11 | res_hidden_size=256, 12 | num_blocks=3, 13 | root_index=0) 14 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/default_runtime.py: -------------------------------------------------------------------------------- 1 | checkpoint_config = dict(interval=10) 2 | 3 | log_config = dict( 4 | interval=50, 5 | hooks=[ 6 | dict(type='TextLoggerHook'), 7 | # dict(type='TensorboardLoggerHook') 8 | ]) 9 | 10 | log_level = 'INFO' 11 | load_from = None 12 | resume_from = None 13 | dist_params = dict(backend='nccl') 14 | workflow = [('train', 1)] 15 | 16 | # disable opencv multithreading to avoid system being overloaded 17 | opencv_num_threads = 0 18 | # set multi-process start method as `fork` to speed up the training 19 | mp_start_method = 'fork' 20 | -------------------------------------------------------------------------------- /configs/examples/pose_estimation/test_camera.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | name='Test Webcam', 4 | camera_id=0, 5 | camera_max_fps=30, 6 | nodes=[ 7 | dict(type='MonitorNode', 8 | name='monitor', 9 | enable_key='m', 10 | enable=False, 11 | input_buffer='_frame_', 12 | output_buffer='display'), 13 | # 'RecorderNode': 14 | # This node save the output video into a file. 15 | dict(type='RecorderNode', 16 | name='recorder', 17 | out_video_file='webcam_output.mp4', 18 | input_buffer='display', 19 | output_buffer='_display_') 20 | ]) 21 | -------------------------------------------------------------------------------- /configs/examples/new_year/README.md: -------------------------------------------------------------------------------- 1 | # New Year Hat and Firecracker Effects 2 | 3 | This demo provides new year effects with pose estimation results, like adding hat on the head and firecracker in the hands. 4 | 5 |
6 |
7 |
8 | 9 | ## Instruction 10 | 11 | ### Get started 12 | 13 | Launch the demo from the mmpose root directory: 14 | 15 | ```shell 16 | python run.py --config configs/examples/new_year/new_year.py 17 | ``` 18 | 19 | ### Hotkeys 20 | 21 | | Hotkey | Function | 22 | | ------ | ------------------------------------- | 23 | | t | Toggle the hat effect on/off. | 24 | | f | Toggle the firecracker effect on/off. | 25 | | h | Show help information. | 26 | | m | Show the monitoring information. | 27 | | q | Exit. | 28 | -------------------------------------------------------------------------------- /configs/examples/valentine_magic/README.md: -------------------------------------------------------------------------------- 1 | # Valentine Magic 2 | 3 | Do you want to show your **love** to your beloved one, especially on Valentine's Day? Express it with your pose using MMPose right away and see the Valentine Magic! 4 | 5 | Try to pose a **hand heart** gesture, and see what will happen? 6 | 7 | Prefer a **blow kiss**? Here comes your flying heart~ 8 | 9 |
10 |
11 |
12 | 13 | ## Instruction 14 | 15 | ### Get started 16 | 17 | Launch the demo from the mmpose root directory: 18 | 19 | ```shell 20 | python run.py --config configs/examples/valentine_magic/valentine_magic.py 21 | ``` 22 | 23 | ### Hotkeys 24 | 25 | | Hotkey | Function | 26 | | ------ | ----------------------------------------- | 27 | | l | Toggle the Valentine Magic effect on/off. | 28 | | v | Toggle the pose visualization on/off. | 29 | | h | Show help information. | 30 | | m | Show diagnostic information. | 31 | | q | Exit. | 32 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/deepfashion_lower.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='deepfashion_lower', 3 | paper_info=dict( 4 | author='Liu, Ziwei and Luo, Ping and Qiu, Shi ' 5 | 'and Wang, Xiaogang and Tang, Xiaoou', 6 | title='DeepFashion: Powering Robust Clothes Recognition ' 7 | 'and Retrieval with Rich Annotations', 8 | container='Proceedings of IEEE Conference on Computer ' 9 | 'Vision and Pattern Recognition (CVPR)', 10 | year='2016', 11 | homepage='http://mmlab.ie.cuhk.edu.hk/projects/' 12 | 'DeepFashion/LandmarkDetection.html', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='left waistline', 17 | id=0, 18 | color=[255, 255, 255], 19 | type='', 20 | swap='right waistline'), 21 | 1: 22 | dict(name='right waistline', 23 | id=1, 24 | color=[255, 255, 255], 25 | type='', 26 | swap='left waistline'), 27 | 2: 28 | dict(name='left hem', 29 | id=2, 30 | color=[255, 255, 255], 31 | type='', 32 | swap='right hem'), 33 | 3: 34 | dict(name='right hem', 35 | id=3, 36 | color=[255, 255, 255], 37 | type='', 38 | swap='left hem'), 39 | }, 40 | skeleton_info={}, 41 | joint_weights=[1.] * 4, 42 | sigmas=[]) 43 | -------------------------------------------------------------------------------- /.pre-commit-config.yaml: -------------------------------------------------------------------------------- 1 | exclude: ^tests/data/ 2 | repos: 3 | - repo: https://github.com/PyCQA/flake8 4 | rev: 3.8.3 5 | hooks: 6 | - id: flake8 7 | - repo: https://github.com/PyCQA/isort 8 | rev: 5.10.1 9 | hooks: 10 | - id: isort 11 | - repo: https://github.com/pre-commit/mirrors-yapf 12 | rev: v0.30.0 13 | hooks: 14 | - id: yapf 15 | - repo: https://github.com/pre-commit/pre-commit-hooks 16 | rev: v3.1.0 17 | hooks: 18 | - id: trailing-whitespace 19 | - id: check-yaml 20 | - id: end-of-file-fixer 21 | - id: requirements-txt-fixer 22 | - id: double-quote-string-fixer 23 | - id: check-merge-conflict 24 | - id: fix-encoding-pragma 25 | args: ["--remove"] 26 | - id: mixed-line-ending 27 | args: ["--fix=lf"] 28 | - repo: https://github.com/myint/docformatter 29 | rev: v1.3.1 30 | hooks: 31 | - id: docformatter 32 | args: ["--in-place", "--wrap-descriptions", "79"] 33 | - repo: https://github.com/codespell-project/codespell 34 | rev: v2.1.0 35 | hooks: 36 | - id: codespell 37 | args: ["--skip", "*.ipynb", "-L", "mot"] 38 | - repo: https://github.com/executablebooks/mdformat 39 | rev: 0.7.14 40 | hooks: 41 | - id: mdformat 42 | args: ["--number"] 43 | additional_dependencies: 44 | - git+https://github.com/mzr1996/mdformat-tables.git#egg=mdformat-tables 45 | - mdformat-gfm 46 | - mdformat_frontmatter 47 | - linkify-it-py 48 | -------------------------------------------------------------------------------- /configs/examples/bing_dwen_dwen/README.md: -------------------------------------------------------------------------------- 1 | # Meow Dwen Dwen 2 | 3 | Do you know [Bing DwenDwen (冰墩墩)](https://en.wikipedia.org/wiki/Bing_Dwen_Dwen_and_Shuey_Rhon_Rhon), the mascot of 2022 Beijing Olympic Games? 4 | 5 |
6 |
7 |
8 | 9 | Now you can dress your cat up in this costume and TA-DA! Be prepared for super cute **Meow Dwen Dwen**. 10 | 11 |
12 |
13 |
14 | 15 | You are a dog fan? Hold on, here comes Woof Dwen Dwen. 16 | 17 |
18 |
19 |
20 | 21 | ## Instruction 22 | 23 | ### Get started 24 | 25 | Launch the demo from the mmpose root directory: 26 | 27 | ```shell 28 | python run.py --config configs/examples/bing_dwen_dwen/bing_dwen_dwen.py 29 | ``` 30 | 31 | ### Hotkeys 32 | 33 | | Hotkey | Function | 34 | | ------ | ---------------------------- | 35 | | s | Change the background. | 36 | | h | Show help information. | 37 | | m | Show diagnostic information. | 38 | | q | Exit. | 39 | 40 | ### Configuration 41 | 42 | - **Use video input** 43 | 44 | As you can see in the config, we set `camera_id` as the path of the input image. You can also set it as a video file path (or url), or a webcam ID number (e.g. `camera_id=0`), to capture the dynamic face from the video input. 45 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/deepfashion_upper.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='deepfashion_upper', 3 | paper_info=dict( 4 | author='Liu, Ziwei and Luo, Ping and Qiu, Shi ' 5 | 'and Wang, Xiaogang and Tang, Xiaoou', 6 | title='DeepFashion: Powering Robust Clothes Recognition ' 7 | 'and Retrieval with Rich Annotations', 8 | container='Proceedings of IEEE Conference on Computer ' 9 | 'Vision and Pattern Recognition (CVPR)', 10 | year='2016', 11 | homepage='http://mmlab.ie.cuhk.edu.hk/projects/' 12 | 'DeepFashion/LandmarkDetection.html', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='left collar', 17 | id=0, 18 | color=[255, 255, 255], 19 | type='', 20 | swap='right collar'), 21 | 1: 22 | dict(name='right collar', 23 | id=1, 24 | color=[255, 255, 255], 25 | type='', 26 | swap='left collar'), 27 | 2: 28 | dict(name='left sleeve', 29 | id=2, 30 | color=[255, 255, 255], 31 | type='', 32 | swap='right sleeve'), 33 | 3: 34 | dict(name='right sleeve', 35 | id=3, 36 | color=[255, 255, 255], 37 | type='', 38 | swap='left sleeve'), 39 | 4: 40 | dict(name='left hem', 41 | id=4, 42 | color=[255, 255, 255], 43 | type='', 44 | swap='right hem'), 45 | 5: 46 | dict(name='right hem', 47 | id=5, 48 | color=[255, 255, 255], 49 | type='', 50 | swap='left hem'), 51 | }, 52 | skeleton_info={}, 53 | joint_weights=[1.] * 6, 54 | sigmas=[]) 55 | -------------------------------------------------------------------------------- /configs/examples/face_swap/README.md: -------------------------------------------------------------------------------- 1 | # Sunglasses and Bug-eye Effects 2 | 3 | Look! Where is my face?:eyes: And whose face is it?:laughing: 4 | 5 |
6 |
7 |
8 | 9 | ## Instruction 10 | 11 | ### Get started 12 | 13 | Launch the demo from the mmpose root directory: 14 | 15 | ```shell 16 | python run.py --config configs/examples/face_swap/face_swap.py 17 | ``` 18 | 19 | ### Hotkeys 20 | 21 | | Hotkey | Function | 22 | | ------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | 23 | | s | Switch between modes | 24 | | v | Toggle the pose visualization on/off. | 25 | | h | Show help information. | 26 | | m | Show diagnostic information. | 27 | | q | Exit. | 28 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # mmpose-webcam-demo 2 | 3 | 简体中文 | [English](/README_en.md) 4 | 5 | 基于 MMPose Webcam API 开发应用或 demo 的项目模板 6 | 7 |
8 | 9 |
10 | 11 | ## 配置环境 12 | 13 | ### 创建虚拟环境 14 | 15 | ```shell 16 | conda create -n mmpose-demo python=3.9 pytorch=1.10 cudatoolkit=11.3 torchvision -c pytorch -y 17 | conda activate mmpose-demo 18 | ``` 19 | 20 | ### 安装 MMCV 和 MMDetection 21 | 22 | ```shell 23 | pip install openmim 24 | mim install mmcv-full 25 | pip install mmdet 26 | ``` 27 | 28 | ### 安装 MMPose 29 | 30 | 为了能随时同步最新的 MMPose 代码，我们推荐将 MMPose 克隆到本地，并通过开发模式安装 31 | 32 | ```shell 33 | cd .. 34 | git clone clone https://github.com/open-mmlab/mmpose.git 35 | cd mmpose 36 | pip install -e . 37 | ``` 38 | 39 | 测试 MMPose 安装成功 40 | 41 | ```shell 42 | python -c "from mmpose.apis import webcam" 43 | ``` 44 | 45 | ### 配置 pre-commit hook 46 | 47 | ```shell 48 | # 在 mmpose-webcam-demo 目录中执行以下操作 49 | pip install pre-commit 50 | pre-commit install 51 | ``` 52 | 53 | ## 运行示例 54 | 55 | ```shell 56 | # use GPU 57 | python run.py --config configs/pose_estimation/pose_estimation.py 58 | 59 | # use CPU 60 | python run.py --config configs/pose_estimation/pose_estimation.py --cpu 61 | 62 | # use debug mode 63 | python run.py --config configs/pose_estimation/pose_estimation.py --debug 64 | ``` 65 | 66 | ## 相关链接 67 | 68 | - 关于摄像头应用接口（MMPose Webcam API） 69 | - [教程文档](https://mmpose.readthedocs.io/zh_CN/latest/tutorials/7_webcam_api.html) 70 | - [API 查询](https://mmpose.readthedocs.io/zh_CN/latest/api.html#mmpose-apis-webcam) 71 | - 关于 MMPose 72 | - [代码仓库](https://github.com/open-mmlab/mmpose) 73 | - [使用文档](https://mmpose.readthedocs.io/zh_CN/latest/) 74 | - [模型池](https://mmpose.readthedocs.io/zh_CN/latest/modelzoo.html) 75 | - 关于 “万物生姿” MMPose 姿态估计创意 Demo 大赛 76 | - [活动主页](https://openmmlab.com/community/mmpose-demo) 77 | - [作品提交](https://github.com/open-mmlab/mmpose/issues/1407) 78 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/deepfashion_full.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='deepfashion_full', 3 | paper_info=dict( 4 | author='Liu, Ziwei and Luo, Ping and Qiu, Shi ' 5 | 'and Wang, Xiaogang and Tang, Xiaoou', 6 | title='DeepFashion: Powering Robust Clothes Recognition ' 7 | 'and Retrieval with Rich Annotations', 8 | container='Proceedings of IEEE Conference on Computer ' 9 | 'Vision and Pattern Recognition (CVPR)', 10 | year='2016', 11 | homepage='http://mmlab.ie.cuhk.edu.hk/projects/' 12 | 'DeepFashion/LandmarkDetection.html', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='left collar', 17 | id=0, 18 | color=[255, 255, 255], 19 | type='', 20 | swap='right collar'), 21 | 1: 22 | dict(name='right collar', 23 | id=1, 24 | color=[255, 255, 255], 25 | type='', 26 | swap='left collar'), 27 | 2: 28 | dict(name='left sleeve', 29 | id=2, 30 | color=[255, 255, 255], 31 | type='', 32 | swap='right sleeve'), 33 | 3: 34 | dict(name='right sleeve', 35 | id=3, 36 | color=[255, 255, 255], 37 | type='', 38 | swap='left sleeve'), 39 | 4: 40 | dict(name='left waistline', 41 | id=0, 42 | color=[255, 255, 255], 43 | type='', 44 | swap='right waistline'), 45 | 5: 46 | dict(name='right waistline', 47 | id=1, 48 | color=[255, 255, 255], 49 | type='', 50 | swap='left waistline'), 51 | 6: 52 | dict(name='left hem', 53 | id=2, 54 | color=[255, 255, 255], 55 | type='', 56 | swap='right hem'), 57 | 7: 58 | dict(name='right hem', 59 | id=3, 60 | color=[255, 255, 255], 61 | type='', 62 | swap='left hem'), 63 | }, 64 | skeleton_info={}, 65 | joint_weights=[1.] * 8, 66 | sigmas=[]) 67 | -------------------------------------------------------------------------------- /README_en.md: -------------------------------------------------------------------------------- 1 | # mmpose-webcam-demo 2 | 3 | [简体中文](/README.md) | English 4 | 5 | A template to build demos and applications with MMPose Webcam API 6 | 7 |
8 | 9 |
10 | 11 | ## Configure Environment 12 | 13 | ### Create a conda environment 14 | 15 | ```shell 16 | conda create -n mmpose-demo python=3.9 pytorch=1.10 cudatoolkit=11.3 torchvision -c pytorch -y 17 | conda activate mmpose-demo 18 | ``` 19 | 20 | ### Install MMCV and MMDetection 21 | 22 | ```shell 23 | pip install openmim 24 | mim install mmcv-full 25 | pip install mmdet 26 | ``` 27 | 28 | ### Install MMPose 29 | 30 | Clone MMPose and install it in the editable mode, so the latest updates of MMPose can be synchronized easily. 31 | 32 | ```shell 33 | cd .. 34 | git clone clone https://github.com/open-mmlab/mmpose.git 35 | cd mmpose 36 | pip install -e . 37 | ``` 38 | 39 | Check MMPose installation. 40 | 41 | ```shell 42 | python -c "from mmpose.apis import webcam" 43 | ``` 44 | 45 | ### Configure pre-commit hook 46 | 47 | ```shell 48 | # In mmpose-webcam-demo repo folder 49 | pip install pre-commit 50 | pre-commit install 51 | ``` 52 | 53 | ## Run the Demo 54 | 55 | ```shell 56 | # use GPU 57 | python run.py --config configs/pose_estimation/pose_estimation.py 58 | 59 | # use CPU 60 | python run.py --config configs/pose_estimation/pose_estimation.py --cpu 61 | 62 | # use debug mode 63 | python run.py --config configs/pose_estimation/pose_estimation.py --debug 64 | ``` 65 | 66 | ## Useful Linnks 67 | 68 | - Webcam API 69 | - [Tutorial](https://mmpose.readthedocs.io/en/latest/tutorials/7_webcam_api.html) 70 | - [API Reference](https://mmpose.readthedocs.io/en/latest/api.html#mmpose-apis-webcam) 71 | - MMPose 72 | - [Code](https://github.com/open-mmlab/mmpose) 73 | - [Documentation](https://mmpose.readthedocs.io/en/latest/) 74 | - [Model Zoo](https://mmpose.readthedocs.io/en/latest/modelzoo.html) 75 | - About "Infinity Pose" MMPose Creative Demo Competition 76 | - [Event Home Page](https://openmmlab.com/community/mmpose-demo) 77 | - [Submission](https://github.com/open-mmlab/mmpose/issues/1407) 78 | -------------------------------------------------------------------------------- /run.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | 3 | import logging 4 | import sys 5 | from argparse import ArgumentParser 6 | 7 | from mmcv import Config, DictAction 8 | from mmpose.apis.webcam import WebcamExecutor 9 | from mmpose.apis.webcam.nodes import model_nodes 10 | 11 | sys.path.append('.') 12 | from src import * # noqa 13 | 14 | 15 | def parse_args(): 16 | parser = ArgumentParser('Webcam executor configs') 17 | parser.add_argument( 18 | '--config', 19 | type=str, 20 | default='configs/examples/pose_estimation/pose_estimation.py') 21 | 22 | parser.add_argument( 23 | '--cfg-options', 24 | nargs='+', 25 | action=DictAction, 26 | default={}, 27 | help='Override settings in the config. The key-value pair ' 28 | 'in xxx=yyy format will be merged into config file. For example, ' 29 | "'--cfg-options executor_cfg.camera_id=1'") 30 | parser.add_argument('--debug', 31 | action='store_true', 32 | help='Show debug information') 33 | 34 | parser.add_argument('--cpu', 35 | action='store_true', 36 | help='Use CPU for model inference.') 37 | parser.add_argument('--cuda', 38 | action='store_true', 39 | help='Use GPU for model inference.') 40 | 41 | return parser.parse_args() 42 | 43 | 44 | def set_device(cfg: Config, device: str): 45 | """Set model device in config. 46 | 47 | Args: 48 | cfg (Config): Webcam config 49 | device (str): device indicator like "cpu" or "cuda:0" 50 | """ 51 | 52 | device = device.lower() 53 | assert device == 'cpu' or device.startswith('cuda:') 54 | 55 | for node_cfg in cfg.executor_cfg.nodes: 56 | if node_cfg.type in model_nodes.__all__: 57 | node_cfg.update(device=device) 58 | 59 | return cfg 60 | 61 | 62 | def run(): 63 | args = parse_args() 64 | cfg = Config.fromfile(args.config) 65 | cfg.merge_from_dict(args.cfg_options) 66 | 67 | if args.debug: 68 | logging.basicConfig(level=logging.DEBUG) 69 | 70 | if args.cpu: 71 | cfg = set_device(cfg, 'cpu') 72 | 73 | if args.cuda: 74 | cfg = set_device(cfg, 'cuda:0') 75 | 76 | webcam_exe = WebcamExecutor(**cfg.executor_cfg) 77 | webcam_exe.run() 78 | 79 | 80 | if __name__ == '__main__': 81 | run() 82 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/zebra.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='zebra', 3 | paper_info=dict( 4 | author='Graving, Jacob M and Chae, Daniel and Naik, Hemal and ' 5 | 'Li, Liang and Koger, Benjamin and Costelloe, Blair R and ' 6 | 'Couzin, Iain D', 7 | title='DeepPoseKit, a software toolkit for fast and robust ' 8 | 'animal pose estimation using deep learning', 9 | container='Elife', 10 | year='2019', 11 | homepage='https://github.com/jgraving/DeepPoseKit-Data', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='snout', id=0, color=[255, 255, 255], type='', swap=''), 16 | 1: 17 | dict(name='head', id=1, color=[255, 255, 255], type='', swap=''), 18 | 2: 19 | dict(name='neck', id=2, color=[255, 255, 255], type='', swap=''), 20 | 3: 21 | dict(name='forelegL1', 22 | id=3, 23 | color=[255, 255, 255], 24 | type='', 25 | swap='forelegR1'), 26 | 4: 27 | dict(name='forelegR1', 28 | id=4, 29 | color=[255, 255, 255], 30 | type='', 31 | swap='forelegL1'), 32 | 5: 33 | dict(name='hindlegL1', 34 | id=5, 35 | color=[255, 255, 255], 36 | type='', 37 | swap='hindlegR1'), 38 | 6: 39 | dict(name='hindlegR1', 40 | id=6, 41 | color=[255, 255, 255], 42 | type='', 43 | swap='hindlegL1'), 44 | 7: 45 | dict(name='tailbase', id=7, color=[255, 255, 255], type='', swap=''), 46 | 8: 47 | dict(name='tailtip', id=8, color=[255, 255, 255], type='', swap='') 48 | }, 49 | skeleton_info={ 50 | 0: dict(link=('head', 'snout'), id=0, color=[255, 255, 255]), 51 | 1: dict(link=('neck', 'head'), id=1, color=[255, 255, 255]), 52 | 2: dict(link=('forelegL1', 'neck'), id=2, color=[255, 255, 255]), 53 | 3: dict(link=('forelegR1', 'neck'), id=3, color=[255, 255, 255]), 54 | 4: dict(link=('hindlegL1', 'tailbase'), id=4, color=[255, 255, 255]), 55 | 5: dict(link=('hindlegR1', 'tailbase'), id=5, color=[255, 255, 255]), 56 | 6: dict(link=('tailbase', 'neck'), id=6, color=[255, 255, 255]), 57 | 7: dict(link=('tailtip', 'tailbase'), id=7, color=[255, 255, 255]) 58 | }, 59 | joint_weights=[1.] * 9, 60 | sigmas=[]) 61 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | pip-wheel-metadata/ 24 | share/python-wheels/ 25 | *.egg-info/ 26 | .installed.cfg 27 | *.egg 28 | MANIFEST 29 | 30 | # PyInstaller 31 | # Usually these files are written by a python script from a template 32 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 33 | *.manifest 34 | *.spec 35 | 36 | # Installer logs 37 | pip-log.txt 38 | pip-delete-this-directory.txt 39 | 40 | # Unit test / coverage reports 41 | htmlcov/ 42 | .tox/ 43 | .nox/ 44 | .coverage 45 | .coverage.* 46 | .cache 47 | nosetests.xml 48 | coverage.xml 49 | *.cover 50 | *.py,cover 51 | .hypothesis/ 52 | .pytest_cache/ 53 | 54 | # Translations 55 | *.mo 56 | *.pot 57 | 58 | # Django stuff: 59 | *.log 60 | local_settings.py 61 | db.sqlite3 62 | db.sqlite3-journal 63 | 64 | # Flask stuff: 65 | instance/ 66 | .webassets-cache 67 | 68 | # Scrapy stuff: 69 | .scrapy 70 | 71 | # Sphinx documentation 72 | docs/_build/ 73 | 74 | # PyBuilder 75 | target/ 76 | 77 | # Jupyter Notebook 78 | .ipynb_checkpoints 79 | 80 | # IPython 81 | profile_default/ 82 | ipython_config.py 83 | 84 | # pyenv 85 | .python-version 86 | 87 | # pipenv 88 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 89 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 90 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 91 | # install all needed dependencies. 92 | #Pipfile.lock 93 | 94 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow 95 | __pypackages__/ 96 | 97 | # Celery stuff 98 | celerybeat-schedule 99 | celerybeat.pid 100 | 101 | # SageMath parsed files 102 | *.sage.py 103 | 104 | # Environments 105 | .env 106 | .venv 107 | env/ 108 | venv/ 109 | ENV/ 110 | env.bak/ 111 | venv.bak/ 112 | 113 | # Spyder project settings 114 | .spyderproject 115 | .spyproject 116 | 117 | # Rope project settings 118 | .ropeproject 119 | 120 | # mkdocs documentation 121 | /site 122 | 123 | # mypy 124 | .mypy_cache/ 125 | .dmypy.json 126 | dmypy.json 127 | 128 | # Pyre type checker 129 | .pyre/ 130 | 131 | # Custom items 132 | *.mp4 133 | -------------------------------------------------------------------------------- /configs/examples/face_swap/face_swap.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | name='FaceSwap', 4 | camera_id=0, 5 | camera_max_fps=20, 6 | nodes=[ 7 | dict( 8 | type='DetectorNode', 9 | name='detector', 10 | model_config='model_configs/mmdet/' 11 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 12 | model_checkpoint='https://download.openmmlab.com' 13 | '/mmdetection/v2.0/ssd/' 14 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 15 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 16 | input_buffer='_input_', # `_input_` is an executor-reserved buffer 17 | output_buffer='det_result'), 18 | dict(type='TopDownPoseEstimatorNode', 19 | name='human pose estimator', 20 | model_config='model_configs/mmpose/' 21 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 22 | model_checkpoint='https://download.openmmlab.com/mmpose/top_down/' 23 | 'vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 24 | '-e2158108_20211205.pth', 25 | labels=['person'], 26 | smooth=True, 27 | input_buffer='det_result', 28 | output_buffer='human_pose'), 29 | dict( 30 | type='ObjectAssignerNode', 31 | name='ResultBinder', 32 | frame_buffer='_frame_', # `_frame_` is a runner-reserved buffer 33 | object_buffer='human_pose', 34 | output_buffer='frame'), 35 | dict(type='FaceSwapNode', 36 | name='FaceSwapper', 37 | mode_key='s', 38 | input_buffer='frame', 39 | output_buffer='face_swap'), 40 | dict(type='ObjectVisualizerNode', 41 | name='Visualizer', 42 | enable_key='v', 43 | input_buffer='face_swap', 44 | output_buffer='vis_pose'), 45 | dict(type='NoticeBoardNode', 46 | name='Help Information', 47 | enable_key='h', 48 | content_lines=[ 49 | 'Swap your faces! ', 50 | 'Hot-keys:', 51 | '"v": Toggle the pose visualization on/off.', 52 | '"s": Switch between modes: Shuffle, Clone and None', 53 | '"h": Show help information', 54 | '"m": Show diagnostic information', 55 | '"q": Exit', 56 | ], 57 | input_buffer='vis_pose', 58 | output_buffer='vis_notice'), 59 | dict(type='MonitorNode', 60 | name='Monitor', 61 | enable_key='m', 62 | enable=False, 63 | input_buffer='vis_notice', 64 | output_buffer='display'), 65 | dict(type='RecorderNode', 66 | name='Recorder', 67 | out_video_file='faceswap_output.mp4', 68 | input_buffer='display', 69 | output_buffer='_display_') 70 | ]) 71 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/aflw.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='aflw', 3 | paper_info=dict( 4 | author='Koestinger, Martin and Wohlhart, Paul and ' 5 | 'Roth, Peter M and Bischof, Horst', 6 | title='Annotated facial landmarks in the wild: ' 7 | 'A large-scale, real-world database for facial ' 8 | 'landmark localization', 9 | container='2011 IEEE international conference on computer ' 10 | 'vision workshops (ICCV workshops)', 11 | year='2011', 12 | homepage='https://www.tugraz.at/institute/icg/research/' 13 | 'team-bischof/lrs/downloads/aflw/', 14 | ), 15 | keypoint_info={ 16 | 0: 17 | dict(name='kpt-0', id=0, color=[255, 255, 255], type='', swap='kpt-5'), 18 | 1: 19 | dict(name='kpt-1', id=1, color=[255, 255, 255], type='', swap='kpt-4'), 20 | 2: 21 | dict(name='kpt-2', id=2, color=[255, 255, 255], type='', swap='kpt-3'), 22 | 3: 23 | dict(name='kpt-3', id=3, color=[255, 255, 255], type='', swap='kpt-2'), 24 | 4: 25 | dict(name='kpt-4', id=4, color=[255, 255, 255], type='', swap='kpt-1'), 26 | 5: 27 | dict(name='kpt-5', id=5, color=[255, 255, 255], type='', swap='kpt-0'), 28 | 6: 29 | dict(name='kpt-6', id=6, color=[255, 255, 255], type='', 30 | swap='kpt-11'), 31 | 7: 32 | dict(name='kpt-7', id=7, color=[255, 255, 255], type='', 33 | swap='kpt-10'), 34 | 8: 35 | dict(name='kpt-8', id=8, color=[255, 255, 255], type='', swap='kpt-9'), 36 | 9: 37 | dict(name='kpt-9', id=9, color=[255, 255, 255], type='', swap='kpt-8'), 38 | 10: 39 | dict(name='kpt-10', 40 | id=10, 41 | color=[255, 255, 255], 42 | type='', 43 | swap='kpt-7'), 44 | 11: 45 | dict(name='kpt-11', 46 | id=11, 47 | color=[255, 255, 255], 48 | type='', 49 | swap='kpt-6'), 50 | 12: 51 | dict(name='kpt-12', 52 | id=12, 53 | color=[255, 255, 255], 54 | type='', 55 | swap='kpt-14'), 56 | 13: 57 | dict(name='kpt-13', id=13, color=[255, 255, 255], type='', swap=''), 58 | 14: 59 | dict(name='kpt-14', 60 | id=14, 61 | color=[255, 255, 255], 62 | type='', 63 | swap='kpt-12'), 64 | 15: 65 | dict(name='kpt-15', 66 | id=15, 67 | color=[255, 255, 255], 68 | type='', 69 | swap='kpt-17'), 70 | 16: 71 | dict(name='kpt-16', id=16, color=[255, 255, 255], type='', swap=''), 72 | 17: 73 | dict(name='kpt-17', 74 | id=17, 75 | color=[255, 255, 255], 76 | type='', 77 | swap='kpt-15'), 78 | 18: 79 | dict(name='kpt-18', id=18, color=[255, 255, 255], type='', swap='') 80 | }, 81 | skeleton_info={}, 82 | joint_weights=[1.] * 19, 83 | sigmas=[]) 84 | -------------------------------------------------------------------------------- /configs/examples/valentine_magic/valentine_magic.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | # Basic configurations of the runner 4 | name='Valentine Magic', 5 | camera_id=0, 6 | nodes=[ 7 | dict( 8 | type='DetectorNode', 9 | name='detector', 10 | model_config='model_configs/mmdet/' 11 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 12 | model_checkpoint='https://download.openmmlab.com' 13 | '/mmdetection/v2.0/ssd/' 14 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 15 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 16 | bbox_thr=0.7, 17 | input_buffer='_input_', # `_input_` is an executor-reserved buffer 18 | output_buffer='det_result'), 19 | dict(type='TopDownPoseEstimatorNode', 20 | name='human pose estimator', 21 | model_config='model_configs/mmpose/' 22 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 23 | model_checkpoint='https://download.openmmlab.com/mmpose/top_down/' 24 | 'vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 25 | '-e2158108_20211205.pth', 26 | labels=['person'], 27 | smooth=True, 28 | input_buffer='det_result', 29 | output_buffer='pose_result'), 30 | dict( 31 | type='ObjectAssignerNode', 32 | name='object assigner', 33 | frame_buffer='_frame_', # `_frame_` is a runner-reserved buffer 34 | object_buffer='pose_result', 35 | output_buffer='frame'), 36 | dict(type='ObjectVisualizerNode', 37 | name='object visualizer', 38 | enable_key='v', 39 | enable=False, 40 | input_buffer='frame', 41 | output_buffer='vis'), 42 | dict( 43 | type='ValentineMagicNode', 44 | name='valentine magic', 45 | enable_key='l', 46 | input_buffer='vis', 47 | output_buffer='vis_heart', 48 | ), 49 | dict( 50 | type='NoticeBoardNode', 51 | name='Helper', 52 | enable_key='h', 53 | enable=False, 54 | input_buffer='vis_heart', 55 | output_buffer='vis_notice', 56 | content_lines=[ 57 | 'This is a demo for pose visualization and simple image ' 58 | 'effects. Have fun!', '', 'Hot-keys:', 59 | '"h": Show help information', '"l": LoveHeart Effect', 60 | '"v": PoseVisualizer', '"m": Show diagnostic information', 61 | '"q": Exit' 62 | ], 63 | ), 64 | dict(type='MonitorNode', 65 | name='monitor', 66 | enable_key='m', 67 | enable=False, 68 | input_buffer='vis_notice', 69 | output_buffer='display'), # `_frame_` is a runner-reserved buffer 70 | dict(type='RecorderNode', 71 | name='recorder', 72 | out_video_file='valentine.mp4', 73 | input_buffer='display', 74 | output_buffer='_display_') 75 | ]) 76 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/filters/smoothnet_h36m.md: -------------------------------------------------------------------------------- 1 | 2 | 3 |
4 | SmoothNet (arXiv'2021) 5 | 6 | ```bibtex 7 | @article{zeng2021smoothnet, 8 | title={SmoothNet: A Plug-and-Play Network for Refining Human Poses in Videos}, 9 | author={Zeng, Ailing and Yang, Lei and Ju, Xuan and Li, Jiefeng and Wang, Jianyi and Xu, Qiang}, 10 | journal={arXiv preprint arXiv:2112.13715}, 11 | year={2021} 12 | } 13 | ``` 14 | 15 |
16 | 17 | 18 | 19 |
20 | Human3.6M (TPAMI'2014) 21 | 22 | ```bibtex 23 | @article{h36m_pami, 24 | author = {Ionescu, Catalin and Papava, Dragos and Olaru, Vlad and Sminchisescu, Cristian}, 25 | title = {Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments}, 26 | journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence}, 27 | publisher = {IEEE Computer Society}, 28 | volume = {36}, 29 | number = {7}, 30 | pages = {1325-1339}, 31 | month = {jul}, 32 | year = {2014} 33 | } 34 | ``` 35 | 36 |
37 | 38 | The following SmoothNet model checkpoints are available for pose smoothing. The table shows the the performance of [SimpleBaseline3D](https://arxiv.org/abs/1705.03098) on [Human3.6M](https://ieeexplore.ieee.org/abstract/document/6682899/) dataset without/with the SmoothNet plugin, and compares the SmoothNet models with 4 different window sizes (8, 16, 32 and 64). The metrics are MPJPE(mm), P-MEJPE(mm) and Acceleration Error (mm/frame^2). 39 | 40 | | Arch | Window Size | MPJPEw/o | MPJPEw | P-MPJPEw/o | P-MPJPEw | AC. Errw/o | AC. Errw | ckpt | 41 | | :----------------------------------- | :---------: | :-----------------: | :---------------: | :-------------------: | :-----------------: | :-------------------: | :-----------------: | :-----------------------------------: | 42 | | [smoothnet_ws8](/configs/_base_/filters/smoothnet_t8_h36m.py) | 8 | 54.48 | 53.15 | 42.20 | 41.32 | 19.18 | 1.87 | [ckpt](https://download.openmmlab.com/mmpose/plugin/smoothnet/smoothnet_ws8_h36m.pth) | 43 | | [smoothnet_ws16](/configs/_base_/filters/smoothnet_t16_h36m.py) | 16 | 54.48 | 52.74 | 42.20 | 41.20 | 19.18 | 1.22 | [ckpt](https://download.openmmlab.com/mmpose/plugin/smoothnet/smoothnet_ws16_h36m.pth) | 44 | | [smoothnet_ws32](/configs/_base_/filters/smoothnet_t32_h36m.py) | 32 | 54.48 | 52.47 | 42.20 | 40.84 | 19.18 | 0.99 | [ckpt](https://download.openmmlab.com/mmpose/plugin/smoothnet/smoothnet_ws32_h36m.pth) | 45 | | [smoothnet_ws64](/configs/_base_/filters/smoothnet_t64_h36m.py) | 64 | 54.48 | 53.37 | 42.20 | 40.77 | 19.18 | 0.92 | [ckpt](https://download.openmmlab.com/mmpose/plugin/smoothnet/smoothnet_ws64_h36m.pth) | 46 | -------------------------------------------------------------------------------- /configs/examples/bing_dwen_dwen/bing_dwen_dwen.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | # Basic configurations of the runner 4 | name='Little fans of 2022 Beijing Winter Olympics', 5 | # Cat image 6 | camera_id='https://user-images.githubusercontent.com/' 7 | '15977946/152932036-b5554cf8-24cf-40d6-a358-35a106013f11.jpeg', 8 | # Dog image 9 | # camera_id='https://user-images.githubusercontent.com/' 10 | # '15977946/152932051-cd280b35-8066-45a0-8f52-657c8631aaba.jpg', 11 | # Camera input 12 | # camera_id=0, 13 | camera_max_fps=20, 14 | nodes=[ 15 | dict( 16 | type='DetectorNode', 17 | name='Detector', 18 | model_config='model_configs/mmdet/' 19 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 20 | model_checkpoint='https://download.openmmlab.com' 21 | '/mmdetection/v2.0/ssd/' 22 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 23 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 24 | input_buffer='_input_', # `_input_` is a runner-reserved buffer 25 | output_buffer='det_result'), 26 | dict(type='TopDownPoseEstimatorNode', 27 | name='Animal Pose Estimator', 28 | model_config='configs/animal/2d_kpt_sview_rgb_img/topdown_heatmap' 29 | '/ap10k/hrnet_w32_ap10k_256x256.py', 30 | model_checkpoint='https://download.openmmlab.com/mmpose/animal/' 31 | 'hrnet/hrnet_w32_ap10k_256x256-18aac840_20211029.pth', 32 | labels=['cat', 'dog'], 33 | smooth=True, 34 | input_buffer='det_result', 35 | output_buffer='animal_pose'), 36 | dict(type='TopDownPoseEstimatorNode', 37 | name='TopDown Pose Estimator', 38 | model_config='model_configs/mmpose/' 39 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 40 | model_checkpoint='https://download.openmmlab.com/mmpose/top_down/' 41 | 'vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 42 | '-e2158108_20211205.pth', 43 | device='cpu', 44 | labels=['person'], 45 | smooth=True, 46 | input_buffer='animal_pose', 47 | output_buffer='human_pose'), 48 | dict( 49 | type='ObjectAssignerNode', 50 | name='ResultBinder', 51 | frame_buffer='_frame_', # `_frame_` is a runner-reserved buffer 52 | object_buffer='human_pose', 53 | output_buffer='frame'), 54 | dict(type='BingDwenDwenNode', 55 | name='XDwenDwen', 56 | mode_key='s', 57 | resource_file='configs/examples/bing_dwen_dwen/' 58 | 'resource-info.json', 59 | out_shape=(480, 480), 60 | input_buffer='frame', 61 | output_buffer='vis'), 62 | dict( 63 | type='NoticeBoardNode', 64 | name='Helper', 65 | enable_key='h', 66 | enable=False, 67 | input_buffer='vis', 68 | output_buffer='vis_notice', 69 | content_lines=[ 70 | 'Let your pet put on a costume of Bing-Dwen-Dwen, ' 71 | 'the mascot of 2022 Beijing Winter Olympics. Have fun!', '', 72 | 'Hot-keys:', '"s": Change the background', 73 | '"h": Show help information', 74 | '"m": Show diagnostic information', '"q": Exit' 75 | ], 76 | ), 77 | dict(type='MonitorNode', 78 | name='Monitor', 79 | enable_key='m', 80 | enable=False, 81 | input_buffer='vis_notice', 82 | output_buffer='display'), 83 | dict(type='RecorderNode', 84 | name='Recorder', 85 | out_video_file='record.mp4', 86 | input_buffer='display', 87 | output_buffer='_display_' 88 | # `_display_` is a runner-reserved buffer 89 | ) 90 | ]) 91 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/cofw.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='cofw', 3 | paper_info=dict( 4 | author='Burgos-Artizzu, Xavier P and Perona, ' 5 | r'Pietro and Doll{\'a}r, Piotr', 6 | title='Robust face landmark estimation under occlusion', 7 | container='Proceedings of the IEEE international ' 8 | 'conference on computer vision', 9 | year='2013', 10 | homepage='http://www.vision.caltech.edu/xpburgos/ICCV13/', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='kpt-0', id=0, color=[255, 255, 255], type='', swap='kpt-1'), 15 | 1: 16 | dict(name='kpt-1', id=1, color=[255, 255, 255], type='', swap='kpt-0'), 17 | 2: 18 | dict(name='kpt-2', id=2, color=[255, 255, 255], type='', swap='kpt-3'), 19 | 3: 20 | dict(name='kpt-3', id=3, color=[255, 255, 255], type='', swap='kpt-2'), 21 | 4: 22 | dict(name='kpt-4', id=4, color=[255, 255, 255], type='', swap='kpt-6'), 23 | 5: 24 | dict(name='kpt-5', id=5, color=[255, 255, 255], type='', swap='kpt-7'), 25 | 6: 26 | dict(name='kpt-6', id=6, color=[255, 255, 255], type='', swap='kpt-4'), 27 | 7: 28 | dict(name='kpt-7', id=7, color=[255, 255, 255], type='', swap='kpt-5'), 29 | 8: 30 | dict(name='kpt-8', id=8, color=[255, 255, 255], type='', swap='kpt-9'), 31 | 9: 32 | dict(name='kpt-9', id=9, color=[255, 255, 255], type='', swap='kpt-8'), 33 | 10: 34 | dict(name='kpt-10', 35 | id=10, 36 | color=[255, 255, 255], 37 | type='', 38 | swap='kpt-11'), 39 | 11: 40 | dict(name='kpt-11', 41 | id=11, 42 | color=[255, 255, 255], 43 | type='', 44 | swap='kpt-10'), 45 | 12: 46 | dict(name='kpt-12', 47 | id=12, 48 | color=[255, 255, 255], 49 | type='', 50 | swap='kpt-14'), 51 | 13: 52 | dict(name='kpt-13', 53 | id=13, 54 | color=[255, 255, 255], 55 | type='', 56 | swap='kpt-15'), 57 | 14: 58 | dict(name='kpt-14', 59 | id=14, 60 | color=[255, 255, 255], 61 | type='', 62 | swap='kpt-12'), 63 | 15: 64 | dict(name='kpt-15', 65 | id=15, 66 | color=[255, 255, 255], 67 | type='', 68 | swap='kpt-13'), 69 | 16: 70 | dict(name='kpt-16', 71 | id=16, 72 | color=[255, 255, 255], 73 | type='', 74 | swap='kpt-17'), 75 | 17: 76 | dict(name='kpt-17', 77 | id=17, 78 | color=[255, 255, 255], 79 | type='', 80 | swap='kpt-16'), 81 | 18: 82 | dict(name='kpt-18', 83 | id=18, 84 | color=[255, 255, 255], 85 | type='', 86 | swap='kpt-19'), 87 | 19: 88 | dict(name='kpt-19', 89 | id=19, 90 | color=[255, 255, 255], 91 | type='', 92 | swap='kpt-18'), 93 | 20: 94 | dict(name='kpt-20', id=20, color=[255, 255, 255], type='', swap=''), 95 | 21: 96 | dict(name='kpt-21', id=21, color=[255, 255, 255], type='', swap=''), 97 | 22: 98 | dict(name='kpt-22', 99 | id=22, 100 | color=[255, 255, 255], 101 | type='', 102 | swap='kpt-23'), 103 | 23: 104 | dict(name='kpt-23', 105 | id=23, 106 | color=[255, 255, 255], 107 | type='', 108 | swap='kpt-22'), 109 | 24: 110 | dict(name='kpt-24', id=24, color=[255, 255, 255], type='', swap=''), 111 | 25: 112 | dict(name='kpt-25', id=25, color=[255, 255, 255], type='', swap=''), 113 | 26: 114 | dict(name='kpt-26', id=26, color=[255, 255, 255], type='', swap=''), 115 | 27: 116 | dict(name='kpt-27', id=27, color=[255, 255, 255], type='', swap=''), 117 | 28: 118 | dict(name='kpt-28', id=28, color=[255, 255, 255], type='', swap='') 119 | }, 120 | skeleton_info={}, 121 | joint_weights=[1.] * 29, 122 | sigmas=[]) 123 | -------------------------------------------------------------------------------- /configs/examples/pose_estimation/pose_tracking.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | # Basic configurations of the executor 4 | name='Pose Estimation', 5 | camera_id=0, 6 | synchronous=False, 7 | # Define nodes. 8 | # The configuration of a node usually includes: 9 | # 1. 'type': Node class name 10 | # 2. 'name': Node name 11 | # 3. I/O buffers (e.g. 'input_buffer', 'output_buffer'): specify the 12 | # input and output buffer names. This may depend on the node class. 13 | # 4. 'enable_key': assign a hot-key to toggle enable/disable this node. 14 | # This may depend on the node class. 15 | # 5. Other class-specific arguments 16 | nodes=[ 17 | # 'PoseTrackerNode': 18 | # This node performs object detection and pose tracking. Object 19 | # detection is performed every several frames. Pose estimation 20 | # is performed for every frame to get the keypoint as well as the 21 | # interval bbox when object detection is not performed. 22 | dict( 23 | type='PoseTrackerNode', 24 | name='pose tracker', 25 | det_model_config='model_configs/mmdet/' 26 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 27 | det_model_checkpoint='https://download.openmmlab.com' 28 | '/mmdetection/v2.0/ssd/' 29 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 30 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 31 | pose_model_config='model_configs/mmpose/' 32 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 33 | pose_model_checkpoint='https://download.openmmlab.com/mmpose/' 34 | 'top_down/vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 35 | '-e2158108_20211205.pth', 36 | det_interval=10, 37 | labels=['person'], 38 | smooth=True, 39 | device='cuda:0', 40 | input_buffer='_input_', # `_input_` is an executor-reserved buffer 41 | output_buffer='human_pose'), 42 | # 'ObjectAssignerNode': 43 | # This node binds the latest model inference result with the current 44 | # frame. (This means the frame image and inference result may be 45 | # asynchronous). 46 | dict( 47 | type='ObjectAssignerNode', 48 | name='object assigner', 49 | frame_buffer='_frame_', # `_frame_` is an executor-reserved buffer 50 | object_buffer='human_pose', 51 | output_buffer='frame'), 52 | # 'ObjectVisualizerNode': 53 | # This node draw the pose visualization result in the frame image. 54 | # Pose results is needed. 55 | dict(type='ObjectVisualizerNode', 56 | name='object visualizer', 57 | enable_key='v', 58 | input_buffer='frame', 59 | output_buffer='vis'), 60 | # 'NoticeBoardNode': 61 | # This node show a notice board with given content, e.g. help 62 | # information. 63 | dict( 64 | type='NoticeBoardNode', 65 | name='instruction', 66 | enable_key='h', 67 | enable=True, 68 | input_buffer='vis', 69 | output_buffer='vis_notice', 70 | content_lines=[ 71 | 'This is a demo for pose visualization and simple image ' 72 | 'effects. Have fun!', '', 'Hot-keys:', 73 | '"v": Pose estimation result visualization', 74 | '"s": Sunglasses effect B-)', '"b": Bug-eye effect 0_0', 75 | '"h": Show help information', 76 | '"m": Show diagnostic information', '"q": Exit' 77 | ], 78 | ), 79 | # 'MonitorNode': 80 | # This node show diagnostic information in the frame image. It can 81 | # be used for debugging or monitoring system resource status. 82 | dict(type='MonitorNode', 83 | name='monitor', 84 | enable_key='m', 85 | enable=False, 86 | input_buffer='vis_notice', 87 | output_buffer='display'), 88 | # 'RecorderNode': 89 | # This node save the output video into a file. 90 | dict(type='RecorderNode', 91 | name='recorder', 92 | out_video_file='webcam_demo.mp4', 93 | input_buffer='display', 94 | output_buffer='_display_' 95 | # `_display_` is an executor-reserved buffer 96 | ) 97 | ]) 98 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/jhmdb.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='jhmdb', 3 | paper_info=dict( 4 | author='H. Jhuang and J. Gall and S. Zuffi and ' 5 | 'C. Schmid and M. J. Black', 6 | title='Towards understanding action recognition', 7 | container='International Conf. on Computer Vision (ICCV)', 8 | year='2013', 9 | homepage='http://jhmdb.is.tue.mpg.de/dataset', 10 | ), 11 | keypoint_info={ 12 | 0: 13 | dict(name='neck', id=0, color=[255, 128, 0], type='upper', swap=''), 14 | 1: 15 | dict(name='belly', id=1, color=[255, 128, 0], type='upper', swap=''), 16 | 2: 17 | dict(name='head', id=2, color=[255, 128, 0], type='upper', swap=''), 18 | 3: 19 | dict(name='right_shoulder', 20 | id=3, 21 | color=[0, 255, 0], 22 | type='upper', 23 | swap='left_shoulder'), 24 | 4: 25 | dict(name='left_shoulder', 26 | id=4, 27 | color=[0, 255, 0], 28 | type='upper', 29 | swap='right_shoulder'), 30 | 5: 31 | dict(name='right_hip', 32 | id=5, 33 | color=[0, 255, 0], 34 | type='lower', 35 | swap='left_hip'), 36 | 6: 37 | dict(name='left_hip', 38 | id=6, 39 | color=[51, 153, 255], 40 | type='lower', 41 | swap='right_hip'), 42 | 7: 43 | dict(name='right_elbow', 44 | id=7, 45 | color=[51, 153, 255], 46 | type='upper', 47 | swap='left_elbow'), 48 | 8: 49 | dict(name='left_elbow', 50 | id=8, 51 | color=[51, 153, 255], 52 | type='upper', 53 | swap='right_elbow'), 54 | 9: 55 | dict(name='right_knee', 56 | id=9, 57 | color=[51, 153, 255], 58 | type='lower', 59 | swap='left_knee'), 60 | 10: 61 | dict(name='left_knee', 62 | id=10, 63 | color=[255, 128, 0], 64 | type='lower', 65 | swap='right_knee'), 66 | 11: 67 | dict(name='right_wrist', 68 | id=11, 69 | color=[255, 128, 0], 70 | type='upper', 71 | swap='left_wrist'), 72 | 12: 73 | dict(name='left_wrist', 74 | id=12, 75 | color=[255, 128, 0], 76 | type='upper', 77 | swap='right_wrist'), 78 | 13: 79 | dict(name='right_ankle', 80 | id=13, 81 | color=[0, 255, 0], 82 | type='lower', 83 | swap='left_ankle'), 84 | 14: 85 | dict(name='left_ankle', 86 | id=14, 87 | color=[0, 255, 0], 88 | type='lower', 89 | swap='right_ankle') 90 | }, 91 | skeleton_info={ 92 | 0: dict(link=('right_ankle', 'right_knee'), id=0, color=[255, 128, 0]), 93 | 1: dict(link=('right_knee', 'right_hip'), id=1, color=[255, 128, 0]), 94 | 2: dict(link=('right_hip', 'belly'), id=2, color=[255, 128, 0]), 95 | 3: dict(link=('belly', 'left_hip'), id=3, color=[0, 255, 0]), 96 | 4: dict(link=('left_hip', 'left_knee'), id=4, color=[0, 255, 0]), 97 | 5: dict(link=('left_knee', 'left_ankle'), id=5, color=[0, 255, 0]), 98 | 6: dict(link=('belly', 'neck'), id=6, color=[51, 153, 255]), 99 | 7: dict(link=('neck', 'head'), id=7, color=[51, 153, 255]), 100 | 8: dict(link=('neck', 'right_shoulder'), id=8, color=[255, 128, 0]), 101 | 9: dict(link=('right_shoulder', 'right_elbow'), 102 | id=9, 103 | color=[255, 128, 0]), 104 | 10: dict(link=('right_elbow', 'right_wrist'), 105 | id=10, 106 | color=[255, 128, 0]), 107 | 11: dict(link=('neck', 'left_shoulder'), id=11, color=[0, 255, 0]), 108 | 12: dict(link=('left_shoulder', 'left_elbow'), 109 | id=12, 110 | color=[0, 255, 0]), 111 | 13: dict(link=('left_elbow', 'left_wrist'), id=13, color=[0, 255, 0]) 112 | }, 113 | joint_weights=[ 114 | 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.2, 1.2, 1.5, 1.5, 1.5, 1.5 115 | ], 116 | # Adapted from COCO dataset. 117 | sigmas=[ 118 | 0.025, 0.107, 0.025, 0.079, 0.079, 0.107, 0.107, 0.072, 0.072, 0.087, 119 | 0.087, 0.062, 0.062, 0.089, 0.089 120 | ]) 121 | -------------------------------------------------------------------------------- /src/nodes/hat_effect_node.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | from itertools import groupby 3 | from typing import List, Optional, Union 4 | 5 | import cv2 6 | import numpy as np 7 | from mmpose.apis.webcam.nodes import NODES, BaseVisualizerNode 8 | from mmpose.apis.webcam.utils import (get_eye_keypoint_ids, 9 | load_image_from_disk_or_url) 10 | 11 | 12 | @NODES.register_module() 13 | class HatEffectNode(BaseVisualizerNode): 14 | def __init__(self, 15 | name: str, 16 | input_buffer: str, 17 | output_buffer: Union[str, List[str]], 18 | enable_key: Optional[Union[str, int]] = None, 19 | src_img_path: Optional[str] = None): 20 | 21 | super().__init__(name, input_buffer, output_buffer, enable_key) 22 | 23 | if src_img_path is None: 24 | # The image attributes to: 25 | # http://616pic.com/sucai/1m9i70p52.html 26 | src_img_path = 'https://user-images.githubusercontent.' \ 27 | 'com/28900607/149766271-2f591c19-9b67-4' \ 28 | 'd92-8f94-c272396ca141.png' 29 | self.src_img = load_image_from_disk_or_url(src_img_path, 30 | cv2.IMREAD_UNCHANGED) 31 | 32 | @staticmethod 33 | def apply_hat_effect(img, 34 | objects, 35 | hat_img, 36 | left_eye_index, 37 | right_eye_index, 38 | kpt_thr=0.5): 39 | """Apply hat effect. 40 | Args: 41 | img (np.ndarray): Image data. 42 | objects (list[dict]): The list of object information containing: 43 | - "keypoints" ([K,3]): keypoint detection result in 44 | [x, y, score] 45 | hat_img (np.ndarray): Hat image with white alpha channel. 46 | left_eye_index (int): Keypoint index of left eye 47 | right_eye_index (int): Keypoint index of right eye 48 | kpt_thr (float): The score threshold of required keypoints. 49 | """ 50 | img_orig = img.copy() 51 | 52 | img = img_orig.copy() 53 | hm, wm = hat_img.shape[:2] 54 | # anchor points in the sunglasses mask 55 | a = 0.3 56 | b = 0.7 57 | pts_src = np.array([[a * wm, a * hm], [a * wm, b * hm], 58 | [b * wm, a * hm], [b * wm, b * hm]], 59 | dtype=np.float32) 60 | 61 | for obj in objects: 62 | kpts = obj['keypoints'] 63 | 64 | if kpts[left_eye_index, 2] < kpt_thr or \ 65 | kpts[right_eye_index, 2] < kpt_thr: 66 | continue 67 | 68 | kpt_leye = kpts[left_eye_index, :2] 69 | kpt_reye = kpts[right_eye_index, :2] 70 | # orthogonal vector to the left-to-right eyes 71 | vo = 0.5 * (kpt_reye - kpt_leye)[::-1] * [-1, 1] 72 | veye = 0.5 * (kpt_reye - kpt_leye) 73 | 74 | # anchor points in the image by eye positions 75 | pts_tar = np.vstack([ 76 | kpt_reye + 1 * veye + 5 * vo, kpt_reye + 1 * veye + 1 * vo, 77 | kpt_leye - 1 * veye + 5 * vo, kpt_leye - 1 * veye + 1 * vo 78 | ]) 79 | 80 | h_mat, _ = cv2.findHomography(pts_src, pts_tar) 81 | patch = cv2.warpPerspective(hat_img, 82 | h_mat, 83 | dsize=(img.shape[1], img.shape[0]), 84 | borderValue=(255, 255, 255)) 85 | # mask the white background area in the patch with a threshold 200 86 | mask = (patch[:, :, -1] > 128) 87 | patch = patch[:, :, :-1] 88 | mask = mask * (cv2.cvtColor(patch, cv2.COLOR_BGR2GRAY) > 30) 89 | mask = mask.astype(np.uint8) 90 | 91 | img = cv2.copyTo(patch, mask, img) 92 | return img 93 | 94 | def draw(self, input_msg): 95 | canvas = input_msg.get_image() 96 | 97 | objects = input_msg.get_objects(lambda x: 'keypoints' in x) 98 | 99 | for model_cfg, object_group in groupby(objects, 100 | lambda x: x['pose_model_cfg']): 101 | left_eye_idx, right_eye_idx = get_eye_keypoint_ids(model_cfg) 102 | 103 | canvas = self.apply_hat_effect(canvas, object_group, self.src_img, 104 | left_eye_idx, right_eye_idx) 105 | return canvas 106 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/atrw.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='atrw', 3 | paper_info=dict( 4 | author='Li, Shuyuan and Li, Jianguo and Tang, Hanlin ' 5 | 'and Qian, Rui and Lin, Weiyao', 6 | title='ATRW: A Benchmark for Amur Tiger ' 7 | 'Re-identification in the Wild', 8 | container='Proceedings of the 28th ACM ' 9 | 'International Conference on Multimedia', 10 | year='2020', 11 | homepage='https://cvwc2019.github.io/challenge.html', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='left_ear', 16 | id=0, 17 | color=[51, 153, 255], 18 | type='upper', 19 | swap='right_ear'), 20 | 1: 21 | dict(name='right_ear', 22 | id=1, 23 | color=[51, 153, 255], 24 | type='upper', 25 | swap='left_ear'), 26 | 2: 27 | dict(name='nose', id=2, color=[51, 153, 255], type='upper', swap=''), 28 | 3: 29 | dict(name='right_shoulder', 30 | id=3, 31 | color=[255, 128, 0], 32 | type='upper', 33 | swap='left_shoulder'), 34 | 4: 35 | dict(name='right_front_paw', 36 | id=4, 37 | color=[255, 128, 0], 38 | type='upper', 39 | swap='left_front_paw'), 40 | 5: 41 | dict(name='left_shoulder', 42 | id=5, 43 | color=[0, 255, 0], 44 | type='upper', 45 | swap='right_shoulder'), 46 | 6: 47 | dict(name='left_front_paw', 48 | id=6, 49 | color=[0, 255, 0], 50 | type='upper', 51 | swap='right_front_paw'), 52 | 7: 53 | dict(name='right_hip', 54 | id=7, 55 | color=[255, 128, 0], 56 | type='lower', 57 | swap='left_hip'), 58 | 8: 59 | dict(name='right_knee', 60 | id=8, 61 | color=[255, 128, 0], 62 | type='lower', 63 | swap='left_knee'), 64 | 9: 65 | dict(name='right_back_paw', 66 | id=9, 67 | color=[255, 128, 0], 68 | type='lower', 69 | swap='left_back_paw'), 70 | 10: 71 | dict(name='left_hip', 72 | id=10, 73 | color=[0, 255, 0], 74 | type='lower', 75 | swap='right_hip'), 76 | 11: 77 | dict(name='left_knee', 78 | id=11, 79 | color=[0, 255, 0], 80 | type='lower', 81 | swap='right_knee'), 82 | 12: 83 | dict(name='left_back_paw', 84 | id=12, 85 | color=[0, 255, 0], 86 | type='lower', 87 | swap='right_back_paw'), 88 | 13: 89 | dict(name='tail', id=13, color=[51, 153, 255], type='lower', swap=''), 90 | 14: 91 | dict(name='center', id=14, color=[51, 153, 255], type='lower', 92 | swap=''), 93 | }, 94 | skeleton_info={ 95 | 0: 96 | dict(link=('left_ear', 'nose'), id=0, color=[51, 153, 255]), 97 | 1: 98 | dict(link=('right_ear', 'nose'), id=1, color=[51, 153, 255]), 99 | 2: 100 | dict(link=('nose', 'center'), id=2, color=[51, 153, 255]), 101 | 3: 102 | dict(link=('left_shoulder', 'left_front_paw'), id=3, color=[0, 255, 103 | 0]), 104 | 4: 105 | dict(link=('left_shoulder', 'center'), id=4, color=[0, 255, 0]), 106 | 5: 107 | dict(link=('right_shoulder', 'right_front_paw'), 108 | id=5, 109 | color=[255, 128, 0]), 110 | 6: 111 | dict(link=('right_shoulder', 'center'), id=6, color=[255, 128, 0]), 112 | 7: 113 | dict(link=('tail', 'center'), id=7, color=[51, 153, 255]), 114 | 8: 115 | dict(link=('right_back_paw', 'right_knee'), id=8, color=[255, 128, 0]), 116 | 9: 117 | dict(link=('right_knee', 'right_hip'), id=9, color=[255, 128, 0]), 118 | 10: 119 | dict(link=('right_hip', 'tail'), id=10, color=[255, 128, 0]), 120 | 11: 121 | dict(link=('left_back_paw', 'left_knee'), id=11, color=[0, 255, 0]), 122 | 12: 123 | dict(link=('left_knee', 'left_hip'), id=12, color=[0, 255, 0]), 124 | 13: 125 | dict(link=('left_hip', 'tail'), id=13, color=[0, 255, 0]), 126 | }, 127 | joint_weights=[1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.], 128 | sigmas=[ 129 | 0.0277, 0.0823, 0.0831, 0.0202, 0.0716, 0.0263, 0.0646, 0.0302, 0.0440, 130 | 0.0316, 0.0333, 0.0547, 0.0263, 0.0683, 0.0539 131 | ]) 132 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/mpi_inf_3dhp.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='mpi_inf_3dhp', 3 | paper_info=dict( 4 | author='ehta, Dushyant and Rhodin, Helge and Casas, Dan and ' 5 | 'Fua, Pascal and Sotnychenko, Oleksandr and Xu, Weipeng and ' 6 | 'Theobalt, Christian', 7 | title='Monocular 3D Human Pose Estimation In The Wild Using Improved ' 8 | 'CNN Supervision', 9 | container='2017 international conference on 3D vision (3DV)', 10 | year='2017', 11 | homepage='http://gvv.mpi-inf.mpg.de/3dhp-dataset', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='head_top', 16 | id=0, 17 | color=[51, 153, 255], 18 | type='upper', 19 | swap=''), 20 | 1: 21 | dict(name='neck', id=1, color=[51, 153, 255], type='upper', swap=''), 22 | 2: 23 | dict(name='right_shoulder', 24 | id=2, 25 | color=[255, 128, 0], 26 | type='upper', 27 | swap='left_shoulder'), 28 | 3: 29 | dict(name='right_elbow', 30 | id=3, 31 | color=[255, 128, 0], 32 | type='upper', 33 | swap='left_elbow'), 34 | 4: 35 | dict(name='right_wrist', 36 | id=4, 37 | color=[255, 128, 0], 38 | type='upper', 39 | swap='left_wrist'), 40 | 5: 41 | dict(name='left_shoulder', 42 | id=5, 43 | color=[0, 255, 0], 44 | type='upper', 45 | swap='right_shoulder'), 46 | 6: 47 | dict(name='left_elbow', 48 | id=6, 49 | color=[0, 255, 0], 50 | type='upper', 51 | swap='right_elbow'), 52 | 7: 53 | dict(name='left_wrist', 54 | id=7, 55 | color=[0, 255, 0], 56 | type='upper', 57 | swap='right_wrist'), 58 | 8: 59 | dict(name='right_hip', 60 | id=8, 61 | color=[255, 128, 0], 62 | type='lower', 63 | swap='left_hip'), 64 | 9: 65 | dict(name='right_knee', 66 | id=9, 67 | color=[255, 128, 0], 68 | type='lower', 69 | swap='left_knee'), 70 | 10: 71 | dict(name='right_ankle', 72 | id=10, 73 | color=[255, 128, 0], 74 | type='lower', 75 | swap='left_ankle'), 76 | 11: 77 | dict(name='left_hip', 78 | id=11, 79 | color=[0, 255, 0], 80 | type='lower', 81 | swap='right_hip'), 82 | 12: 83 | dict(name='left_knee', 84 | id=12, 85 | color=[0, 255, 0], 86 | type='lower', 87 | swap='right_knee'), 88 | 13: 89 | dict(name='left_ankle', 90 | id=13, 91 | color=[0, 255, 0], 92 | type='lower', 93 | swap='right_ankle'), 94 | 14: 95 | dict(name='root', id=14, color=[51, 153, 255], type='lower', swap=''), 96 | 15: 97 | dict(name='spine', id=15, color=[51, 153, 255], type='upper', swap=''), 98 | 16: 99 | dict(name='head', id=16, color=[51, 153, 255], type='upper', swap='') 100 | }, 101 | skeleton_info={ 102 | 0: dict(link=('neck', 'right_shoulder'), id=0, color=[255, 128, 0]), 103 | 1: dict(link=('right_shoulder', 'right_elbow'), 104 | id=1, 105 | color=[255, 128, 0]), 106 | 2: dict(link=('right_elbow', 'right_wrist'), id=2, color=[255, 128, 107 | 0]), 108 | 3: dict(link=('neck', 'left_shoulder'), id=3, color=[0, 255, 0]), 109 | 4: dict(link=('left_shoulder', 'left_elbow'), id=4, color=[0, 255, 0]), 110 | 5: dict(link=('left_elbow', 'left_wrist'), id=5, color=[0, 255, 0]), 111 | 6: dict(link=('root', 'right_hip'), id=6, color=[255, 128, 0]), 112 | 7: dict(link=('right_hip', 'right_knee'), id=7, color=[255, 128, 0]), 113 | 8: dict(link=('right_knee', 'right_ankle'), id=8, color=[255, 128, 0]), 114 | 9: dict(link=('root', 'left_hip'), id=9, color=[0, 255, 0]), 115 | 10: dict(link=('left_hip', 'left_knee'), id=10, color=[0, 255, 0]), 116 | 11: dict(link=('left_knee', 'left_ankle'), id=11, color=[0, 255, 0]), 117 | 12: dict(link=('head_top', 'head'), id=12, color=[51, 153, 255]), 118 | 13: dict(link=('head', 'neck'), id=13, color=[51, 153, 255]), 119 | 14: dict(link=('neck', 'spine'), id=14, color=[51, 153, 255]), 120 | 15: dict(link=('spine', 'root'), id=15, color=[51, 153, 255]) 121 | }, 122 | joint_weights=[1.] * 17, 123 | sigmas=[]) 124 | -------------------------------------------------------------------------------- /configs/examples/new_year/new_year.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | # Basic configurations of the runner 4 | name='New Year', 5 | camera_id=0, 6 | camera_max_fps=20, 7 | synchronous=False, 8 | # Define nodes. 9 | # The configuration of a node usually includes: 10 | # 1. 'type': Node class name 11 | # 2. 'name': Node name 12 | # 3. I/O buffers (e.g. 'input_buffer', 'output_buffer'): specify the 13 | # input and output buffer names. This may depend on the node class. 14 | # 4. 'enable_key': assign a hot-key to toggle enable/disable this node. 15 | # This may depend on the node class. 16 | # 5. Other class-specific arguments 17 | nodes=[ 18 | dict( 19 | type='DetectorNode', 20 | name='detector', 21 | model_config='model_configs/mmdet/' 22 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 23 | model_checkpoint='https://download.openmmlab.com' 24 | '/mmdetection/v2.0/ssd/' 25 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 26 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 27 | input_buffer='_input_', # `_input_` is an executor-reserved buffer 28 | output_buffer='det_result'), 29 | dict(type='TopDownPoseEstimatorNode', 30 | name='human pose estimator', 31 | model_config='model_configs/mmpose/' 32 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 33 | model_checkpoint='https://download.openmmlab.com/mmpose/top_down/' 34 | 'vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 35 | '-e2158108_20211205.pth', 36 | labels=['person'], 37 | smooth=True, 38 | input_buffer='det_result', 39 | output_buffer='human_pose'), 40 | dict(type='TopDownPoseEstimatorNode', 41 | name='animal pose estimator', 42 | model_config='model_configs/' 43 | 'mmpose/hrnet_w32_animalpose_256x256.py', 44 | model_checkpoint='https://download.openmmlab.com/mmpose/animal/' 45 | 'hrnet/hrnet_w32_animalpose_256x256-1aa7f075_20210426.pth', 46 | labels=['cat', 'dog', 'horse', 'sheep', 'cow'], 47 | input_buffer='human_pose', 48 | output_buffer='animal_pose'), 49 | dict( 50 | type='ObjectAssignerNode', 51 | name='object assigner', 52 | frame_buffer='_frame_', # `_frame_` is an executor-reserved buffer 53 | object_buffer='animal_pose', 54 | output_buffer='frame'), 55 | # 'HatNode': 56 | # This node draw the hat effect in the frame image. 57 | # Pose results is needed. 58 | dict(type='HatEffectNode', 59 | name='hat effect', 60 | enable_key='t', 61 | input_buffer='frame', 62 | output_buffer='vis_hat'), 63 | # 'FirecrackerNode': 64 | # This node draw the firecracker effect in the frame image. 65 | # Pose results is needed. 66 | dict(type='FirecrackerEffectNode', 67 | name='firecracker effect', 68 | enable_key='f', 69 | input_buffer='vis_hat', 70 | output_buffer='vis_firecracker'), 71 | # 'NoticeBoardNode': 72 | # This node show a notice board with given content, e.g. help 73 | # information. 74 | dict( 75 | type='NoticeBoardNode', 76 | name='instruction', 77 | enable_key='h', 78 | enable=False, 79 | input_buffer='vis_firecracker', 80 | output_buffer='vis_notice', 81 | content_lines=[ 82 | 'This is a demo for pose visualization and simple image ' 83 | 'effects. Have fun!', '', 'Hot-keys:', '"t": Hat effect', 84 | '"f": Firecracker effect', '"h": Show help information', 85 | '"m": Show diagnostic information', '"q": Exit' 86 | ], 87 | ), 88 | # 'MonitorNode': 89 | # This node show diagnostic information in the frame image. It can 90 | # be used for debugging or monitoring system resource status. 91 | dict(type='MonitorNode', 92 | name='monitor', 93 | enable_key='m', 94 | enable=False, 95 | input_buffer='vis_notice', 96 | output_buffer='display'), 97 | # 'RecorderNode': 98 | # This node save the output video into a file. 99 | dict(type='RecorderNode', 100 | name='recorder', 101 | out_video_file='new_year.mp4', 102 | input_buffer='display', 103 | output_buffer='_display_' 104 | # `_display_` is a runner-reserved buffer 105 | ) 106 | ]) 107 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/campus.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='campus', 3 | paper_info=dict( 4 | author='Belagiannis, Vasileios and Amin, Sikandar and Andriluka, ' 5 | 'Mykhaylo and Schiele, Bernt and Navab, Nassir and Ilic, Slobodan', 6 | title='3D Pictorial Structures for Multiple Human Pose Estimation', 7 | container='IEEE Computer Society Conference on Computer Vision and ' 8 | 'Pattern Recognition (CVPR)', 9 | year='2014', 10 | homepage='http://campar.in.tum.de/Chair/MultiHumanPose', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='right_ankle', 15 | id=0, 16 | color=[255, 128, 0], 17 | type='lower', 18 | swap='left_ankle'), 19 | 1: 20 | dict(name='right_knee', 21 | id=1, 22 | color=[255, 128, 0], 23 | type='lower', 24 | swap='left_knee'), 25 | 2: 26 | dict(name='right_hip', 27 | id=2, 28 | color=[255, 128, 0], 29 | type='lower', 30 | swap='left_hip'), 31 | 3: 32 | dict(name='left_hip', 33 | id=3, 34 | color=[0, 255, 0], 35 | type='lower', 36 | swap='right_hip'), 37 | 4: 38 | dict(name='left_knee', 39 | id=4, 40 | color=[0, 255, 0], 41 | type='lower', 42 | swap='right_knee'), 43 | 5: 44 | dict(name='left_ankle', 45 | id=5, 46 | color=[0, 255, 0], 47 | type='lower', 48 | swap='right_ankle'), 49 | 6: 50 | dict(name='right_wrist', 51 | id=6, 52 | color=[255, 128, 0], 53 | type='upper', 54 | swap='left_wrist'), 55 | 7: 56 | dict(name='right_elbow', 57 | id=7, 58 | color=[255, 128, 0], 59 | type='upper', 60 | swap='left_elbow'), 61 | 8: 62 | dict(name='right_shoulder', 63 | id=8, 64 | color=[255, 128, 0], 65 | type='upper', 66 | swap='left_shoulder'), 67 | 9: 68 | dict(name='left_shoulder', 69 | id=9, 70 | color=[0, 255, 0], 71 | type='upper', 72 | swap='right_shoulder'), 73 | 10: 74 | dict(name='left_elbow', 75 | id=10, 76 | color=[0, 255, 0], 77 | type='upper', 78 | swap='right_elbow'), 79 | 11: 80 | dict(name='left_wrist', 81 | id=11, 82 | color=[0, 255, 0], 83 | type='upper', 84 | swap='right_wrist'), 85 | 12: 86 | dict(name='bottom_head', 87 | id=12, 88 | color=[51, 153, 255], 89 | type='upper', 90 | swap=''), 91 | 13: 92 | dict(name='top_head', 93 | id=13, 94 | color=[51, 153, 255], 95 | type='upper', 96 | swap=''), 97 | }, 98 | skeleton_info={ 99 | 0: 100 | dict(link=('right_ankle', 'right_knee'), id=0, color=[255, 128, 0]), 101 | 1: 102 | dict(link=('right_knee', 'right_hip'), id=1, color=[255, 128, 0]), 103 | 2: 104 | dict(link=('left_hip', 'left_knee'), id=2, color=[0, 255, 0]), 105 | 3: 106 | dict(link=('left_knee', 'left_ankle'), id=3, color=[0, 255, 0]), 107 | 4: 108 | dict(link=('right_hip', 'left_hip'), id=4, color=[51, 153, 255]), 109 | 5: 110 | dict(link=('right_wrist', 'right_elbow'), id=5, color=[255, 128, 0]), 111 | 6: 112 | dict(link=('right_elbow', 'right_shoulder'), id=6, color=[255, 128, 113 | 0]), 114 | 7: 115 | dict(link=('left_shoulder', 'left_elbow'), id=7, color=[0, 255, 0]), 116 | 8: 117 | dict(link=('left_elbow', 'left_wrist'), id=8, color=[0, 255, 0]), 118 | 9: 119 | dict(link=('right_hip', 'right_shoulder'), id=9, color=[255, 128, 0]), 120 | 10: 121 | dict(link=('left_hip', 'left_shoulder'), id=10, color=[0, 255, 0]), 122 | 11: 123 | dict(link=('right_shoulder', 'bottom_head'), 124 | id=11, 125 | color=[255, 128, 0]), 126 | 12: 127 | dict(link=('left_shoulder', 'bottom_head'), id=12, color=[0, 255, 0]), 128 | 13: 129 | dict(link=('bottom_head', 'top_head'), id=13, color=[51, 153, 255]), 130 | }, 131 | joint_weights=[ 132 | 1.5, 1.2, 1.0, 1.0, 1.2, 1.5, 1.5, 1.2, 1.0, 1.0, 1.2, 1.5, 1.0, 1.0 133 | ], 134 | sigmas=[ 135 | 0.089, 0.087, 0.107, 0.107, 0.087, 0.089, 0.062, 0.072, 0.079, 0.079, 136 | 0.072, 0.062, 0.026, 0.026 137 | ]) 138 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/shelf.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='shelf', 3 | paper_info=dict( 4 | author='Belagiannis, Vasileios and Amin, Sikandar and Andriluka, ' 5 | 'Mykhaylo and Schiele, Bernt and Navab, Nassir and Ilic, Slobodan', 6 | title='3D Pictorial Structures for Multiple Human Pose Estimation', 7 | container='IEEE Computer Society Conference on Computer Vision and ' 8 | 'Pattern Recognition (CVPR)', 9 | year='2014', 10 | homepage='http://campar.in.tum.de/Chair/MultiHumanPose', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='right_ankle', 15 | id=0, 16 | color=[255, 128, 0], 17 | type='lower', 18 | swap='left_ankle'), 19 | 1: 20 | dict(name='right_knee', 21 | id=1, 22 | color=[255, 128, 0], 23 | type='lower', 24 | swap='left_knee'), 25 | 2: 26 | dict(name='right_hip', 27 | id=2, 28 | color=[255, 128, 0], 29 | type='lower', 30 | swap='left_hip'), 31 | 3: 32 | dict(name='left_hip', 33 | id=3, 34 | color=[0, 255, 0], 35 | type='lower', 36 | swap='right_hip'), 37 | 4: 38 | dict(name='left_knee', 39 | id=4, 40 | color=[0, 255, 0], 41 | type='lower', 42 | swap='right_knee'), 43 | 5: 44 | dict(name='left_ankle', 45 | id=5, 46 | color=[0, 255, 0], 47 | type='lower', 48 | swap='right_ankle'), 49 | 6: 50 | dict(name='right_wrist', 51 | id=6, 52 | color=[255, 128, 0], 53 | type='upper', 54 | swap='left_wrist'), 55 | 7: 56 | dict(name='right_elbow', 57 | id=7, 58 | color=[255, 128, 0], 59 | type='upper', 60 | swap='left_elbow'), 61 | 8: 62 | dict(name='right_shoulder', 63 | id=8, 64 | color=[255, 128, 0], 65 | type='upper', 66 | swap='left_shoulder'), 67 | 9: 68 | dict(name='left_shoulder', 69 | id=9, 70 | color=[0, 255, 0], 71 | type='upper', 72 | swap='right_shoulder'), 73 | 10: 74 | dict(name='left_elbow', 75 | id=10, 76 | color=[0, 255, 0], 77 | type='upper', 78 | swap='right_elbow'), 79 | 11: 80 | dict(name='left_wrist', 81 | id=11, 82 | color=[0, 255, 0], 83 | type='upper', 84 | swap='right_wrist'), 85 | 12: 86 | dict(name='bottom_head', 87 | id=12, 88 | color=[51, 153, 255], 89 | type='upper', 90 | swap=''), 91 | 13: 92 | dict(name='top_head', 93 | id=13, 94 | color=[51, 153, 255], 95 | type='upper', 96 | swap=''), 97 | }, 98 | skeleton_info={ 99 | 0: 100 | dict(link=('right_ankle', 'right_knee'), id=0, color=[255, 128, 0]), 101 | 1: 102 | dict(link=('right_knee', 'right_hip'), id=1, color=[255, 128, 0]), 103 | 2: 104 | dict(link=('left_hip', 'left_knee'), id=2, color=[0, 255, 0]), 105 | 3: 106 | dict(link=('left_knee', 'left_ankle'), id=3, color=[0, 255, 0]), 107 | 4: 108 | dict(link=('right_hip', 'left_hip'), id=4, color=[51, 153, 255]), 109 | 5: 110 | dict(link=('right_wrist', 'right_elbow'), id=5, color=[255, 128, 0]), 111 | 6: 112 | dict(link=('right_elbow', 'right_shoulder'), id=6, color=[255, 128, 113 | 0]), 114 | 7: 115 | dict(link=('left_shoulder', 'left_elbow'), id=7, color=[0, 255, 0]), 116 | 8: 117 | dict(link=('left_elbow', 'left_wrist'), id=8, color=[0, 255, 0]), 118 | 9: 119 | dict(link=('right_hip', 'right_shoulder'), id=9, color=[255, 128, 0]), 120 | 10: 121 | dict(link=('left_hip', 'left_shoulder'), id=10, color=[0, 255, 0]), 122 | 11: 123 | dict(link=('right_shoulder', 'bottom_head'), 124 | id=11, 125 | color=[255, 128, 0]), 126 | 12: 127 | dict(link=('left_shoulder', 'bottom_head'), id=12, color=[0, 255, 0]), 128 | 13: 129 | dict(link=('bottom_head', 'top_head'), id=13, color=[51, 153, 255]), 130 | }, 131 | joint_weights=[ 132 | 1.5, 1.2, 1.0, 1.0, 1.2, 1.5, 1.5, 1.2, 1.0, 1.0, 1.2, 1.5, 1.0, 1.0 133 | ], 134 | sigmas=[ 135 | 0.089, 0.087, 0.107, 0.107, 0.087, 0.089, 0.062, 0.072, 0.079, 0.079, 136 | 0.072, 0.062, 0.026, 0.026 137 | ]) 138 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/crowdpose.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='crowdpose', 3 | paper_info=dict( 4 | author='Li, Jiefeng and Wang, Can and Zhu, Hao and ' 5 | 'Mao, Yihuan and Fang, Hao-Shu and Lu, Cewu', 6 | title='CrowdPose: Efficient Crowded Scenes Pose Estimation ' 7 | 'and A New Benchmark', 8 | container='Proceedings of IEEE Conference on Computer ' 9 | 'Vision and Pattern Recognition (CVPR)', 10 | year='2019', 11 | homepage='https://github.com/Jeff-sjtu/CrowdPose', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='left_shoulder', 16 | id=0, 17 | color=[51, 153, 255], 18 | type='upper', 19 | swap='right_shoulder'), 20 | 1: 21 | dict(name='right_shoulder', 22 | id=1, 23 | color=[51, 153, 255], 24 | type='upper', 25 | swap='left_shoulder'), 26 | 2: 27 | dict(name='left_elbow', 28 | id=2, 29 | color=[51, 153, 255], 30 | type='upper', 31 | swap='right_elbow'), 32 | 3: 33 | dict(name='right_elbow', 34 | id=3, 35 | color=[51, 153, 255], 36 | type='upper', 37 | swap='left_elbow'), 38 | 4: 39 | dict(name='left_wrist', 40 | id=4, 41 | color=[51, 153, 255], 42 | type='upper', 43 | swap='right_wrist'), 44 | 5: 45 | dict(name='right_wrist', 46 | id=5, 47 | color=[0, 255, 0], 48 | type='upper', 49 | swap='left_wrist'), 50 | 6: 51 | dict(name='left_hip', 52 | id=6, 53 | color=[255, 128, 0], 54 | type='lower', 55 | swap='right_hip'), 56 | 7: 57 | dict(name='right_hip', 58 | id=7, 59 | color=[0, 255, 0], 60 | type='lower', 61 | swap='left_hip'), 62 | 8: 63 | dict(name='left_knee', 64 | id=8, 65 | color=[255, 128, 0], 66 | type='lower', 67 | swap='right_knee'), 68 | 9: 69 | dict(name='right_knee', 70 | id=9, 71 | color=[0, 255, 0], 72 | type='lower', 73 | swap='left_knee'), 74 | 10: 75 | dict(name='left_ankle', 76 | id=10, 77 | color=[255, 128, 0], 78 | type='lower', 79 | swap='right_ankle'), 80 | 11: 81 | dict(name='right_ankle', 82 | id=11, 83 | color=[0, 255, 0], 84 | type='lower', 85 | swap='left_ankle'), 86 | 12: 87 | dict(name='top_head', 88 | id=12, 89 | color=[255, 128, 0], 90 | type='upper', 91 | swap=''), 92 | 13: 93 | dict(name='neck', id=13, color=[0, 255, 0], type='upper', swap='') 94 | }, 95 | skeleton_info={ 96 | 0: 97 | dict(link=('left_ankle', 'left_knee'), id=0, color=[0, 255, 0]), 98 | 1: 99 | dict(link=('left_knee', 'left_hip'), id=1, color=[0, 255, 0]), 100 | 2: 101 | dict(link=('right_ankle', 'right_knee'), id=2, color=[255, 128, 0]), 102 | 3: 103 | dict(link=('right_knee', 'right_hip'), id=3, color=[255, 128, 0]), 104 | 4: 105 | dict(link=('left_hip', 'right_hip'), id=4, color=[51, 153, 255]), 106 | 5: 107 | dict(link=('left_shoulder', 'left_hip'), id=5, color=[51, 153, 255]), 108 | 6: 109 | dict(link=('right_shoulder', 'right_hip'), id=6, color=[51, 153, 255]), 110 | 7: 111 | dict(link=('left_shoulder', 'right_shoulder'), 112 | id=7, 113 | color=[51, 153, 255]), 114 | 8: 115 | dict(link=('left_shoulder', 'left_elbow'), id=8, color=[0, 255, 0]), 116 | 9: 117 | dict(link=('right_shoulder', 'right_elbow'), id=9, color=[255, 128, 118 | 0]), 119 | 10: 120 | dict(link=('left_elbow', 'left_wrist'), id=10, color=[0, 255, 0]), 121 | 11: 122 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 123 | 12: 124 | dict(link=('top_head', 'neck'), id=12, color=[51, 153, 255]), 125 | 13: 126 | dict(link=('right_shoulder', 'neck'), id=13, color=[51, 153, 255]), 127 | 14: 128 | dict(link=('left_shoulder', 'neck'), id=14, color=[51, 153, 255]) 129 | }, 130 | joint_weights=[ 131 | 0.2, 0.2, 0.2, 1.3, 1.5, 0.2, 1.3, 1.5, 0.2, 0.2, 0.5, 0.2, 0.2, 0.5 132 | ], 133 | sigmas=[ 134 | 0.079, 0.079, 0.072, 0.072, 0.062, 0.062, 0.107, 0.107, 0.087, 0.087, 135 | 0.089, 0.089, 0.079, 0.079 136 | ]) 137 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/aic.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='aic', 3 | paper_info=dict( 4 | author='Wu, Jiahong and Zheng, He and Zhao, Bo and ' 5 | 'Li, Yixin and Yan, Baoming and Liang, Rui and ' 6 | 'Wang, Wenjia and Zhou, Shipei and Lin, Guosen and ' 7 | 'Fu, Yanwei and others', 8 | title='Ai challenger: A large-scale dataset for going ' 9 | 'deeper in image understanding', 10 | container='arXiv', 11 | year='2017', 12 | homepage='https://github.com/AIChallenger/AI_Challenger_2017', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='right_shoulder', 17 | id=0, 18 | color=[255, 128, 0], 19 | type='upper', 20 | swap='left_shoulder'), 21 | 1: 22 | dict(name='right_elbow', 23 | id=1, 24 | color=[255, 128, 0], 25 | type='upper', 26 | swap='left_elbow'), 27 | 2: 28 | dict(name='right_wrist', 29 | id=2, 30 | color=[255, 128, 0], 31 | type='upper', 32 | swap='left_wrist'), 33 | 3: 34 | dict(name='left_shoulder', 35 | id=3, 36 | color=[0, 255, 0], 37 | type='upper', 38 | swap='right_shoulder'), 39 | 4: 40 | dict(name='left_elbow', 41 | id=4, 42 | color=[0, 255, 0], 43 | type='upper', 44 | swap='right_elbow'), 45 | 5: 46 | dict(name='left_wrist', 47 | id=5, 48 | color=[0, 255, 0], 49 | type='upper', 50 | swap='right_wrist'), 51 | 6: 52 | dict(name='right_hip', 53 | id=6, 54 | color=[255, 128, 0], 55 | type='lower', 56 | swap='left_hip'), 57 | 7: 58 | dict(name='right_knee', 59 | id=7, 60 | color=[255, 128, 0], 61 | type='lower', 62 | swap='left_knee'), 63 | 8: 64 | dict(name='right_ankle', 65 | id=8, 66 | color=[255, 128, 0], 67 | type='lower', 68 | swap='left_ankle'), 69 | 9: 70 | dict(name='left_hip', 71 | id=9, 72 | color=[0, 255, 0], 73 | type='lower', 74 | swap='right_hip'), 75 | 10: 76 | dict(name='left_knee', 77 | id=10, 78 | color=[0, 255, 0], 79 | type='lower', 80 | swap='right_knee'), 81 | 11: 82 | dict(name='left_ankle', 83 | id=11, 84 | color=[0, 255, 0], 85 | type='lower', 86 | swap='right_ankle'), 87 | 12: 88 | dict(name='head_top', 89 | id=12, 90 | color=[51, 153, 255], 91 | type='upper', 92 | swap=''), 93 | 13: 94 | dict(name='neck', id=13, color=[51, 153, 255], type='upper', swap='') 95 | }, 96 | skeleton_info={ 97 | 0: dict(link=('right_wrist', 'right_elbow'), id=0, color=[255, 128, 98 | 0]), 99 | 1: dict(link=('right_elbow', 'right_shoulder'), 100 | id=1, 101 | color=[255, 128, 0]), 102 | 2: dict(link=('right_shoulder', 'neck'), id=2, color=[51, 153, 255]), 103 | 3: dict(link=('neck', 'left_shoulder'), id=3, color=[51, 153, 255]), 104 | 4: dict(link=('left_shoulder', 'left_elbow'), id=4, color=[0, 255, 0]), 105 | 5: dict(link=('left_elbow', 'left_wrist'), id=5, color=[0, 255, 0]), 106 | 6: dict(link=('right_ankle', 'right_knee'), id=6, color=[255, 128, 0]), 107 | 7: dict(link=('right_knee', 'right_hip'), id=7, color=[255, 128, 0]), 108 | 8: dict(link=('right_hip', 'left_hip'), id=8, color=[51, 153, 255]), 109 | 9: dict(link=('left_hip', 'left_knee'), id=9, color=[0, 255, 0]), 110 | 10: dict(link=('left_knee', 'left_ankle'), id=10, color=[0, 255, 0]), 111 | 11: dict(link=('head_top', 'neck'), id=11, color=[51, 153, 255]), 112 | 12: dict(link=('right_shoulder', 'right_hip'), 113 | id=12, 114 | color=[51, 153, 255]), 115 | 13: dict(link=('left_shoulder', 'left_hip'), 116 | id=13, 117 | color=[51, 153, 255]) 118 | }, 119 | joint_weights=[ 120 | 1., 1.2, 1.5, 1., 1.2, 1.5, 1., 1.2, 1.5, 1., 1.2, 1.5, 1., 1. 121 | ], 122 | 123 | # 'https://github.com/AIChallenger/AI_Challenger_2017/blob/master/' 124 | # 'Evaluation/keypoint_eval/keypoint_eval.py#L50' 125 | # delta = 2 x sigma 126 | sigmas=[ 127 | 0.01388152, 0.01515228, 0.01057665, 0.01417709, 0.01497891, 0.01402144, 128 | 0.03909642, 0.03686941, 0.01981803, 0.03843971, 0.03412318, 0.02415081, 129 | 0.01291456, 0.01236173 130 | ]) 131 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/ap10k.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='ap10k', 3 | paper_info=dict( 4 | author='Yu, Hang and Xu, Yufei and Zhang, Jing and ' 5 | 'Zhao, Wei and Guan, Ziyu and Tao, Dacheng', 6 | title='AP-10K: A Benchmark for Animal Pose Estimation in the Wild', 7 | container='35th Conference on Neural Information Processing Systems ' 8 | '(NeurIPS 2021) Track on Datasets and Bench-marks.', 9 | year='2021', 10 | homepage='https://github.com/AlexTheBad/AP-10K', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='L_Eye', id=0, color=[0, 255, 0], type='upper', 15 | swap='R_Eye'), 16 | 1: 17 | dict(name='R_Eye', 18 | id=1, 19 | color=[255, 128, 0], 20 | type='upper', 21 | swap='L_Eye'), 22 | 2: 23 | dict(name='Nose', id=2, color=[51, 153, 255], type='upper', swap=''), 24 | 3: 25 | dict(name='Neck', id=3, color=[51, 153, 255], type='upper', swap=''), 26 | 4: 27 | dict(name='Root of tail', 28 | id=4, 29 | color=[51, 153, 255], 30 | type='lower', 31 | swap=''), 32 | 5: 33 | dict(name='L_Shoulder', 34 | id=5, 35 | color=[51, 153, 255], 36 | type='upper', 37 | swap='R_Shoulder'), 38 | 6: 39 | dict(name='L_Elbow', 40 | id=6, 41 | color=[51, 153, 255], 42 | type='upper', 43 | swap='R_Elbow'), 44 | 7: 45 | dict(name='L_F_Paw', 46 | id=7, 47 | color=[0, 255, 0], 48 | type='upper', 49 | swap='R_F_Paw'), 50 | 8: 51 | dict(name='R_Shoulder', 52 | id=8, 53 | color=[0, 255, 0], 54 | type='upper', 55 | swap='L_Shoulder'), 56 | 9: 57 | dict(name='R_Elbow', 58 | id=9, 59 | color=[255, 128, 0], 60 | type='upper', 61 | swap='L_Elbow'), 62 | 10: 63 | dict(name='R_F_Paw', 64 | id=10, 65 | color=[0, 255, 0], 66 | type='lower', 67 | swap='L_F_Paw'), 68 | 11: 69 | dict(name='L_Hip', 70 | id=11, 71 | color=[255, 128, 0], 72 | type='lower', 73 | swap='R_Hip'), 74 | 12: 75 | dict(name='L_Knee', 76 | id=12, 77 | color=[255, 128, 0], 78 | type='lower', 79 | swap='R_Knee'), 80 | 13: 81 | dict(name='L_B_Paw', 82 | id=13, 83 | color=[0, 255, 0], 84 | type='lower', 85 | swap='R_B_Paw'), 86 | 14: 87 | dict(name='R_Hip', 88 | id=14, 89 | color=[0, 255, 0], 90 | type='lower', 91 | swap='L_Hip'), 92 | 15: 93 | dict(name='R_Knee', 94 | id=15, 95 | color=[0, 255, 0], 96 | type='lower', 97 | swap='L_Knee'), 98 | 16: 99 | dict(name='R_B_Paw', 100 | id=16, 101 | color=[0, 255, 0], 102 | type='lower', 103 | swap='L_B_Paw'), 104 | }, 105 | skeleton_info={ 106 | 0: dict(link=('L_Eye', 'R_Eye'), id=0, color=[0, 0, 255]), 107 | 1: dict(link=('L_Eye', 'Nose'), id=1, color=[0, 0, 255]), 108 | 2: dict(link=('R_Eye', 'Nose'), id=2, color=[0, 0, 255]), 109 | 3: dict(link=('Nose', 'Neck'), id=3, color=[0, 255, 0]), 110 | 4: dict(link=('Neck', 'Root of tail'), id=4, color=[0, 255, 0]), 111 | 5: dict(link=('Neck', 'L_Shoulder'), id=5, color=[0, 255, 255]), 112 | 6: dict(link=('L_Shoulder', 'L_Elbow'), id=6, color=[0, 255, 255]), 113 | 7: dict(link=('L_Elbow', 'L_F_Paw'), id=6, color=[0, 255, 255]), 114 | 8: dict(link=('Neck', 'R_Shoulder'), id=7, color=[6, 156, 250]), 115 | 9: dict(link=('R_Shoulder', 'R_Elbow'), id=8, color=[6, 156, 250]), 116 | 10: dict(link=('R_Elbow', 'R_F_Paw'), id=9, color=[6, 156, 250]), 117 | 11: dict(link=('Root of tail', 'L_Hip'), id=10, color=[0, 255, 255]), 118 | 12: dict(link=('L_Hip', 'L_Knee'), id=11, color=[0, 255, 255]), 119 | 13: dict(link=('L_Knee', 'L_B_Paw'), id=12, color=[0, 255, 255]), 120 | 14: dict(link=('Root of tail', 'R_Hip'), id=13, color=[6, 156, 250]), 121 | 15: dict(link=('R_Hip', 'R_Knee'), id=14, color=[6, 156, 250]), 122 | 16: dict(link=('R_Knee', 'R_B_Paw'), id=15, color=[6, 156, 250]), 123 | }, 124 | joint_weights=[ 125 | 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.5, 1.5, 1., 1., 1.2, 1.2, 1.5, 126 | 1.5 127 | ], 128 | sigmas=[ 129 | 0.025, 0.025, 0.026, 0.035, 0.035, 0.079, 0.072, 0.062, 0.079, 0.072, 130 | 0.062, 0.107, 0.087, 0.089, 0.107, 0.087, 0.089 131 | ]) 132 | -------------------------------------------------------------------------------- /model_configs/mmpose/vipnas_mbv3_coco_wholebody_256x192_dark.py: -------------------------------------------------------------------------------- 1 | _base_ = ['./_base_/default_runtime.py', './_base_/datasets/coco_wholebody.py'] 2 | evaluation = dict(interval=10, metric='mAP', save_best='AP') 3 | 4 | optimizer = dict( 5 | type='Adam', 6 | lr=5e-4, 7 | ) 8 | optimizer_config = dict(grad_clip=None) 9 | # learning policy 10 | lr_config = dict(policy='step', 11 | warmup='linear', 12 | warmup_iters=500, 13 | warmup_ratio=0.001, 14 | step=[170, 200]) 15 | total_epochs = 210 16 | channel_cfg = dict(num_output_channels=133, 17 | dataset_joints=133, 18 | dataset_channel=[ 19 | list(range(133)), 20 | ], 21 | inference_channel=list(range(133))) 22 | 23 | # model settings 24 | model = dict(type='TopDown', 25 | pretrained=None, 26 | backbone=dict(type='ViPNAS_MobileNetV3'), 27 | keypoint_head=dict( 28 | type='ViPNASHeatmapSimpleHead', 29 | in_channels=160, 30 | out_channels=channel_cfg['num_output_channels'], 31 | num_deconv_filters=(160, 160, 160), 32 | num_deconv_groups=(160, 160, 160), 33 | loss_keypoint=dict(type='JointsMSELoss', 34 | use_target_weight=True)), 35 | train_cfg=dict(), 36 | test_cfg=dict(flip_test=True, 37 | post_process='unbiased', 38 | shift_heatmap=True, 39 | modulate_kernel=11)) 40 | 41 | data_cfg = dict( 42 | image_size=[192, 256], 43 | heatmap_size=[48, 64], 44 | num_output_channels=channel_cfg['num_output_channels'], 45 | num_joints=channel_cfg['dataset_joints'], 46 | dataset_channel=channel_cfg['dataset_channel'], 47 | inference_channel=channel_cfg['inference_channel'], 48 | soft_nms=False, 49 | nms_thr=1.0, 50 | oks_thr=0.9, 51 | vis_thr=0.2, 52 | use_gt_bbox=False, 53 | det_bbox_thr=0.0, 54 | bbox_file='data/coco/person_detection_results/' 55 | 'COCO_val2017_detections_AP_H_56_person.json', 56 | ) 57 | 58 | train_pipeline = [ 59 | dict(type='LoadImageFromFile'), 60 | dict(type='TopDownGetBboxCenterScale', padding=1.25), 61 | dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3), 62 | dict(type='TopDownRandomFlip', flip_prob=0.5), 63 | dict(type='TopDownHalfBodyTransform', 64 | num_joints_half_body=8, 65 | prob_half_body=0.3), 66 | dict(type='TopDownGetRandomScaleRotation', 67 | rot_factor=30, 68 | scale_factor=0.25), 69 | dict(type='TopDownAffine'), 70 | dict(type='ToTensor'), 71 | dict(type='NormalizeTensor', 72 | mean=[0.485, 0.456, 0.406], 73 | std=[0.229, 0.224, 0.225]), 74 | dict(type='TopDownGenerateTarget', sigma=2, unbiased_encoding=True), 75 | dict(type='Collect', 76 | keys=['img', 'target', 'target_weight'], 77 | meta_keys=[ 78 | 'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale', 79 | 'rotation', 'bbox_score', 'flip_pairs' 80 | ]), 81 | ] 82 | 83 | val_pipeline = [ 84 | dict(type='LoadImageFromFile'), 85 | dict(type='TopDownGetBboxCenterScale', padding=1.25), 86 | dict(type='TopDownAffine'), 87 | dict(type='ToTensor'), 88 | dict(type='NormalizeTensor', 89 | mean=[0.485, 0.456, 0.406], 90 | std=[0.229, 0.224, 0.225]), 91 | dict(type='Collect', 92 | keys=['img'], 93 | meta_keys=[ 94 | 'image_file', 'center', 'scale', 'rotation', 'bbox_score', 95 | 'flip_pairs' 96 | ]), 97 | ] 98 | 99 | test_pipeline = val_pipeline 100 | 101 | data_root = 'data/coco' 102 | data = dict( 103 | samples_per_gpu=64, 104 | workers_per_gpu=2, 105 | val_dataloader=dict(samples_per_gpu=32), 106 | test_dataloader=dict(samples_per_gpu=32), 107 | train=dict( 108 | type='TopDownCocoWholeBodyDataset', 109 | ann_file=f'{data_root}/annotations/coco_wholebody_train_v1.0.json', 110 | img_prefix=f'{data_root}/train2017/', 111 | data_cfg=data_cfg, 112 | pipeline=train_pipeline, 113 | dataset_info={{_base_.dataset_info}}), 114 | val=dict(type='TopDownCocoWholeBodyDataset', 115 | ann_file=f'{data_root}/annotations/coco_wholebody_val_v1.0.json', 116 | img_prefix=f'{data_root}/val2017/', 117 | data_cfg=data_cfg, 118 | pipeline=val_pipeline, 119 | dataset_info={{_base_.dataset_info}}), 120 | test=dict(type='TopDownCocoWholeBodyDataset', 121 | ann_file=f'{data_root}/annotations/coco_wholebody_val_v1.0.json', 122 | img_prefix=f'{data_root}/val2017/', 123 | data_cfg=data_cfg, 124 | pipeline=test_pipeline, 125 | dataset_info={{_base_.dataset_info}}), 126 | ) 127 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/h36m.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='h36m', 3 | paper_info=dict( 4 | author='Ionescu, Catalin and Papava, Dragos and ' 5 | 'Olaru, Vlad and Sminchisescu, Cristian', 6 | title='Human3.6M: Large Scale Datasets and Predictive ' 7 | 'Methods for 3D Human Sensing in Natural Environments', 8 | container='IEEE Transactions on Pattern Analysis and ' 9 | 'Machine Intelligence', 10 | year='2014', 11 | homepage='http://vision.imar.ro/human3.6m/description.php', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='root', id=0, color=[51, 153, 255], type='lower', swap=''), 16 | 1: 17 | dict(name='right_hip', 18 | id=1, 19 | color=[255, 128, 0], 20 | type='lower', 21 | swap='left_hip'), 22 | 2: 23 | dict(name='right_knee', 24 | id=2, 25 | color=[255, 128, 0], 26 | type='lower', 27 | swap='left_knee'), 28 | 3: 29 | dict(name='right_foot', 30 | id=3, 31 | color=[255, 128, 0], 32 | type='lower', 33 | swap='left_foot'), 34 | 4: 35 | dict(name='left_hip', 36 | id=4, 37 | color=[0, 255, 0], 38 | type='lower', 39 | swap='right_hip'), 40 | 5: 41 | dict(name='left_knee', 42 | id=5, 43 | color=[0, 255, 0], 44 | type='lower', 45 | swap='right_knee'), 46 | 6: 47 | dict(name='left_foot', 48 | id=6, 49 | color=[0, 255, 0], 50 | type='lower', 51 | swap='right_foot'), 52 | 7: 53 | dict(name='spine', id=7, color=[51, 153, 255], type='upper', swap=''), 54 | 8: 55 | dict(name='thorax', id=8, color=[51, 153, 255], type='upper', swap=''), 56 | 9: 57 | dict(name='neck_base', 58 | id=9, 59 | color=[51, 153, 255], 60 | type='upper', 61 | swap=''), 62 | 10: 63 | dict(name='head', id=10, color=[51, 153, 255], type='upper', swap=''), 64 | 11: 65 | dict(name='left_shoulder', 66 | id=11, 67 | color=[0, 255, 0], 68 | type='upper', 69 | swap='right_shoulder'), 70 | 12: 71 | dict(name='left_elbow', 72 | id=12, 73 | color=[0, 255, 0], 74 | type='upper', 75 | swap='right_elbow'), 76 | 13: 77 | dict(name='left_wrist', 78 | id=13, 79 | color=[0, 255, 0], 80 | type='upper', 81 | swap='right_wrist'), 82 | 14: 83 | dict(name='right_shoulder', 84 | id=14, 85 | color=[255, 128, 0], 86 | type='upper', 87 | swap='left_shoulder'), 88 | 15: 89 | dict(name='right_elbow', 90 | id=15, 91 | color=[255, 128, 0], 92 | type='upper', 93 | swap='left_elbow'), 94 | 16: 95 | dict(name='right_wrist', 96 | id=16, 97 | color=[255, 128, 0], 98 | type='upper', 99 | swap='left_wrist') 100 | }, 101 | skeleton_info={ 102 | 0: 103 | dict(link=('root', 'left_hip'), id=0, color=[0, 255, 0]), 104 | 1: 105 | dict(link=('left_hip', 'left_knee'), id=1, color=[0, 255, 0]), 106 | 2: 107 | dict(link=('left_knee', 'left_foot'), id=2, color=[0, 255, 0]), 108 | 3: 109 | dict(link=('root', 'right_hip'), id=3, color=[255, 128, 0]), 110 | 4: 111 | dict(link=('right_hip', 'right_knee'), id=4, color=[255, 128, 0]), 112 | 5: 113 | dict(link=('right_knee', 'right_foot'), id=5, color=[255, 128, 0]), 114 | 6: 115 | dict(link=('root', 'spine'), id=6, color=[51, 153, 255]), 116 | 7: 117 | dict(link=('spine', 'thorax'), id=7, color=[51, 153, 255]), 118 | 8: 119 | dict(link=('thorax', 'neck_base'), id=8, color=[51, 153, 255]), 120 | 9: 121 | dict(link=('neck_base', 'head'), id=9, color=[51, 153, 255]), 122 | 10: 123 | dict(link=('thorax', 'left_shoulder'), id=10, color=[0, 255, 0]), 124 | 11: 125 | dict(link=('left_shoulder', 'left_elbow'), id=11, color=[0, 255, 0]), 126 | 12: 127 | dict(link=('left_elbow', 'left_wrist'), id=12, color=[0, 255, 0]), 128 | 13: 129 | dict(link=('thorax', 'right_shoulder'), id=13, color=[255, 128, 0]), 130 | 14: 131 | dict(link=('right_shoulder', 'right_elbow'), 132 | id=14, 133 | color=[255, 128, 0]), 134 | 15: 135 | dict(link=('right_elbow', 'right_wrist'), id=15, color=[255, 128, 0]) 136 | }, 137 | joint_weights=[1.] * 17, 138 | sigmas=[], 139 | stats_info=dict(bbox_center=(528., 427.), bbox_scale=400.)) 140 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/mpii.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='mpii', 3 | paper_info=dict( 4 | author='Mykhaylo Andriluka and Leonid Pishchulin and ' 5 | 'Peter Gehler and Schiele, Bernt', 6 | title='2D Human Pose Estimation: New Benchmark and ' 7 | 'State of the Art Analysis', 8 | container='IEEE Conference on Computer Vision and ' 9 | 'Pattern Recognition (CVPR)', 10 | year='2014', 11 | homepage='http://human-pose.mpi-inf.mpg.de/', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='right_ankle', 16 | id=0, 17 | color=[255, 128, 0], 18 | type='lower', 19 | swap='left_ankle'), 20 | 1: 21 | dict(name='right_knee', 22 | id=1, 23 | color=[255, 128, 0], 24 | type='lower', 25 | swap='left_knee'), 26 | 2: 27 | dict(name='right_hip', 28 | id=2, 29 | color=[255, 128, 0], 30 | type='lower', 31 | swap='left_hip'), 32 | 3: 33 | dict(name='left_hip', 34 | id=3, 35 | color=[0, 255, 0], 36 | type='lower', 37 | swap='right_hip'), 38 | 4: 39 | dict(name='left_knee', 40 | id=4, 41 | color=[0, 255, 0], 42 | type='lower', 43 | swap='right_knee'), 44 | 5: 45 | dict(name='left_ankle', 46 | id=5, 47 | color=[0, 255, 0], 48 | type='lower', 49 | swap='right_ankle'), 50 | 6: 51 | dict(name='pelvis', id=6, color=[51, 153, 255], type='lower', swap=''), 52 | 7: 53 | dict(name='thorax', id=7, color=[51, 153, 255], type='upper', swap=''), 54 | 8: 55 | dict(name='upper_neck', 56 | id=8, 57 | color=[51, 153, 255], 58 | type='upper', 59 | swap=''), 60 | 9: 61 | dict(name='head_top', 62 | id=9, 63 | color=[51, 153, 255], 64 | type='upper', 65 | swap=''), 66 | 10: 67 | dict(name='right_wrist', 68 | id=10, 69 | color=[255, 128, 0], 70 | type='upper', 71 | swap='left_wrist'), 72 | 11: 73 | dict(name='right_elbow', 74 | id=11, 75 | color=[255, 128, 0], 76 | type='upper', 77 | swap='left_elbow'), 78 | 12: 79 | dict(name='right_shoulder', 80 | id=12, 81 | color=[255, 128, 0], 82 | type='upper', 83 | swap='left_shoulder'), 84 | 13: 85 | dict(name='left_shoulder', 86 | id=13, 87 | color=[0, 255, 0], 88 | type='upper', 89 | swap='right_shoulder'), 90 | 14: 91 | dict(name='left_elbow', 92 | id=14, 93 | color=[0, 255, 0], 94 | type='upper', 95 | swap='right_elbow'), 96 | 15: 97 | dict(name='left_wrist', 98 | id=15, 99 | color=[0, 255, 0], 100 | type='upper', 101 | swap='right_wrist') 102 | }, 103 | skeleton_info={ 104 | 0: 105 | dict(link=('right_ankle', 'right_knee'), id=0, color=[255, 128, 0]), 106 | 1: 107 | dict(link=('right_knee', 'right_hip'), id=1, color=[255, 128, 0]), 108 | 2: 109 | dict(link=('right_hip', 'pelvis'), id=2, color=[255, 128, 0]), 110 | 3: 111 | dict(link=('pelvis', 'left_hip'), id=3, color=[0, 255, 0]), 112 | 4: 113 | dict(link=('left_hip', 'left_knee'), id=4, color=[0, 255, 0]), 114 | 5: 115 | dict(link=('left_knee', 'left_ankle'), id=5, color=[0, 255, 0]), 116 | 6: 117 | dict(link=('pelvis', 'thorax'), id=6, color=[51, 153, 255]), 118 | 7: 119 | dict(link=('thorax', 'upper_neck'), id=7, color=[51, 153, 255]), 120 | 8: 121 | dict(link=('upper_neck', 'head_top'), id=8, color=[51, 153, 255]), 122 | 9: 123 | dict(link=('upper_neck', 'right_shoulder'), id=9, color=[255, 128, 0]), 124 | 10: 125 | dict(link=('right_shoulder', 'right_elbow'), 126 | id=10, 127 | color=[255, 128, 0]), 128 | 11: 129 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 130 | 12: 131 | dict(link=('upper_neck', 'left_shoulder'), id=12, color=[0, 255, 0]), 132 | 13: 133 | dict(link=('left_shoulder', 'left_elbow'), id=13, color=[0, 255, 0]), 134 | 14: 135 | dict(link=('left_elbow', 'left_wrist'), id=14, color=[0, 255, 0]) 136 | }, 137 | joint_weights=[ 138 | 1.5, 1.2, 1., 1., 1.2, 1.5, 1., 1., 1., 1., 1.5, 1.2, 1., 1., 1.2, 1.5 139 | ], 140 | # Adapted from COCO dataset. 141 | sigmas=[ 142 | 0.089, 0.083, 0.107, 0.107, 0.083, 0.089, 0.026, 0.026, 0.026, 0.026, 143 | 0.062, 0.072, 0.179, 0.179, 0.072, 0.062 144 | ]) 145 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/mhp.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='mhp', 3 | paper_info=dict( 4 | author='Zhao, Jian and Li, Jianshu and Cheng, Yu and ' 5 | 'Sim, Terence and Yan, Shuicheng and Feng, Jiashi', 6 | title='Understanding humans in crowded scenes: ' 7 | 'Deep nested adversarial learning and a ' 8 | 'new benchmark for multi-human parsing', 9 | container='Proceedings of the 26th ACM ' 10 | 'international conference on Multimedia', 11 | year='2018', 12 | homepage='https://lv-mhp.github.io/dataset', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='right_ankle', 17 | id=0, 18 | color=[255, 128, 0], 19 | type='lower', 20 | swap='left_ankle'), 21 | 1: 22 | dict(name='right_knee', 23 | id=1, 24 | color=[255, 128, 0], 25 | type='lower', 26 | swap='left_knee'), 27 | 2: 28 | dict(name='right_hip', 29 | id=2, 30 | color=[255, 128, 0], 31 | type='lower', 32 | swap='left_hip'), 33 | 3: 34 | dict(name='left_hip', 35 | id=3, 36 | color=[0, 255, 0], 37 | type='lower', 38 | swap='right_hip'), 39 | 4: 40 | dict(name='left_knee', 41 | id=4, 42 | color=[0, 255, 0], 43 | type='lower', 44 | swap='right_knee'), 45 | 5: 46 | dict(name='left_ankle', 47 | id=5, 48 | color=[0, 255, 0], 49 | type='lower', 50 | swap='right_ankle'), 51 | 6: 52 | dict(name='pelvis', id=6, color=[51, 153, 255], type='lower', swap=''), 53 | 7: 54 | dict(name='thorax', id=7, color=[51, 153, 255], type='upper', swap=''), 55 | 8: 56 | dict(name='upper_neck', 57 | id=8, 58 | color=[51, 153, 255], 59 | type='upper', 60 | swap=''), 61 | 9: 62 | dict(name='head_top', 63 | id=9, 64 | color=[51, 153, 255], 65 | type='upper', 66 | swap=''), 67 | 10: 68 | dict(name='right_wrist', 69 | id=10, 70 | color=[255, 128, 0], 71 | type='upper', 72 | swap='left_wrist'), 73 | 11: 74 | dict(name='right_elbow', 75 | id=11, 76 | color=[255, 128, 0], 77 | type='upper', 78 | swap='left_elbow'), 79 | 12: 80 | dict(name='right_shoulder', 81 | id=12, 82 | color=[255, 128, 0], 83 | type='upper', 84 | swap='left_shoulder'), 85 | 13: 86 | dict(name='left_shoulder', 87 | id=13, 88 | color=[0, 255, 0], 89 | type='upper', 90 | swap='right_shoulder'), 91 | 14: 92 | dict(name='left_elbow', 93 | id=14, 94 | color=[0, 255, 0], 95 | type='upper', 96 | swap='right_elbow'), 97 | 15: 98 | dict(name='left_wrist', 99 | id=15, 100 | color=[0, 255, 0], 101 | type='upper', 102 | swap='right_wrist') 103 | }, 104 | skeleton_info={ 105 | 0: 106 | dict(link=('right_ankle', 'right_knee'), id=0, color=[255, 128, 0]), 107 | 1: 108 | dict(link=('right_knee', 'right_hip'), id=1, color=[255, 128, 0]), 109 | 2: 110 | dict(link=('right_hip', 'pelvis'), id=2, color=[255, 128, 0]), 111 | 3: 112 | dict(link=('pelvis', 'left_hip'), id=3, color=[0, 255, 0]), 113 | 4: 114 | dict(link=('left_hip', 'left_knee'), id=4, color=[0, 255, 0]), 115 | 5: 116 | dict(link=('left_knee', 'left_ankle'), id=5, color=[0, 255, 0]), 117 | 6: 118 | dict(link=('pelvis', 'thorax'), id=6, color=[51, 153, 255]), 119 | 7: 120 | dict(link=('thorax', 'upper_neck'), id=7, color=[51, 153, 255]), 121 | 8: 122 | dict(link=('upper_neck', 'head_top'), id=8, color=[51, 153, 255]), 123 | 9: 124 | dict(link=('upper_neck', 'right_shoulder'), id=9, color=[255, 128, 0]), 125 | 10: 126 | dict(link=('right_shoulder', 'right_elbow'), 127 | id=10, 128 | color=[255, 128, 0]), 129 | 11: 130 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 131 | 12: 132 | dict(link=('upper_neck', 'left_shoulder'), id=12, color=[0, 255, 0]), 133 | 13: 134 | dict(link=('left_shoulder', 'left_elbow'), id=13, color=[0, 255, 0]), 135 | 14: 136 | dict(link=('left_elbow', 'left_wrist'), id=14, color=[0, 255, 0]) 137 | }, 138 | joint_weights=[ 139 | 1.5, 1.2, 1., 1., 1.2, 1.5, 1., 1., 1., 1., 1.5, 1.2, 1., 1., 1.2, 1.5 140 | ], 141 | # Adapted from COCO dataset. 142 | sigmas=[ 143 | 0.089, 0.083, 0.107, 0.107, 0.083, 0.089, 0.026, 0.026, 0.026, 0.026, 144 | 0.062, 0.072, 0.179, 0.179, 0.072, 0.062 145 | ]) 146 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/interhand2d.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='interhand2d', 3 | paper_info=dict( 4 | author='Moon, Gyeongsik and Yu, Shoou-I and Wen, He and ' 5 | 'Shiratori, Takaaki and Lee, Kyoung Mu', 6 | title='InterHand2.6M: A dataset and baseline for 3D ' 7 | 'interacting hand pose estimation from a single RGB image', 8 | container='arXiv', 9 | year='2020', 10 | homepage='https://mks0601.github.io/InterHand2.6M/', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='thumb4', id=0, color=[255, 128, 0], type='', swap=''), 15 | 1: 16 | dict(name='thumb3', id=1, color=[255, 128, 0], type='', swap=''), 17 | 2: 18 | dict(name='thumb2', id=2, color=[255, 128, 0], type='', swap=''), 19 | 3: 20 | dict(name='thumb1', id=3, color=[255, 128, 0], type='', swap=''), 21 | 4: 22 | dict(name='forefinger4', id=4, color=[255, 153, 255], type='', 23 | swap=''), 24 | 5: 25 | dict(name='forefinger3', id=5, color=[255, 153, 255], type='', 26 | swap=''), 27 | 6: 28 | dict(name='forefinger2', id=6, color=[255, 153, 255], type='', 29 | swap=''), 30 | 7: 31 | dict(name='forefinger1', id=7, color=[255, 153, 255], type='', 32 | swap=''), 33 | 8: 34 | dict(name='middle_finger4', 35 | id=8, 36 | color=[102, 178, 255], 37 | type='', 38 | swap=''), 39 | 9: 40 | dict(name='middle_finger3', 41 | id=9, 42 | color=[102, 178, 255], 43 | type='', 44 | swap=''), 45 | 10: 46 | dict(name='middle_finger2', 47 | id=10, 48 | color=[102, 178, 255], 49 | type='', 50 | swap=''), 51 | 11: 52 | dict(name='middle_finger1', 53 | id=11, 54 | color=[102, 178, 255], 55 | type='', 56 | swap=''), 57 | 12: 58 | dict(name='ring_finger4', id=12, color=[255, 51, 51], type='', 59 | swap=''), 60 | 13: 61 | dict(name='ring_finger3', id=13, color=[255, 51, 51], type='', 62 | swap=''), 63 | 14: 64 | dict(name='ring_finger2', id=14, color=[255, 51, 51], type='', 65 | swap=''), 66 | 15: 67 | dict(name='ring_finger1', id=15, color=[255, 51, 51], type='', 68 | swap=''), 69 | 16: 70 | dict(name='pinky_finger4', id=16, color=[0, 255, 0], type='', swap=''), 71 | 17: 72 | dict(name='pinky_finger3', id=17, color=[0, 255, 0], type='', swap=''), 73 | 18: 74 | dict(name='pinky_finger2', id=18, color=[0, 255, 0], type='', swap=''), 75 | 19: 76 | dict(name='pinky_finger1', id=19, color=[0, 255, 0], type='', swap=''), 77 | 20: 78 | dict(name='wrist', id=20, color=[255, 255, 255], type='', swap='') 79 | }, 80 | skeleton_info={ 81 | 0: 82 | dict(link=('wrist', 'thumb1'), id=0, color=[255, 128, 0]), 83 | 1: 84 | dict(link=('thumb1', 'thumb2'), id=1, color=[255, 128, 0]), 85 | 2: 86 | dict(link=('thumb2', 'thumb3'), id=2, color=[255, 128, 0]), 87 | 3: 88 | dict(link=('thumb3', 'thumb4'), id=3, color=[255, 128, 0]), 89 | 4: 90 | dict(link=('wrist', 'forefinger1'), id=4, color=[255, 153, 255]), 91 | 5: 92 | dict(link=('forefinger1', 'forefinger2'), id=5, color=[255, 153, 255]), 93 | 6: 94 | dict(link=('forefinger2', 'forefinger3'), id=6, color=[255, 153, 255]), 95 | 7: 96 | dict(link=('forefinger3', 'forefinger4'), id=7, color=[255, 153, 255]), 97 | 8: 98 | dict(link=('wrist', 'middle_finger1'), id=8, color=[102, 178, 255]), 99 | 9: 100 | dict(link=('middle_finger1', 'middle_finger2'), 101 | id=9, 102 | color=[102, 178, 255]), 103 | 10: 104 | dict(link=('middle_finger2', 'middle_finger3'), 105 | id=10, 106 | color=[102, 178, 255]), 107 | 11: 108 | dict(link=('middle_finger3', 'middle_finger4'), 109 | id=11, 110 | color=[102, 178, 255]), 111 | 12: 112 | dict(link=('wrist', 'ring_finger1'), id=12, color=[255, 51, 51]), 113 | 13: 114 | dict(link=('ring_finger1', 'ring_finger2'), id=13, color=[255, 51, 115 | 51]), 116 | 14: 117 | dict(link=('ring_finger2', 'ring_finger3'), id=14, color=[255, 51, 118 | 51]), 119 | 15: 120 | dict(link=('ring_finger3', 'ring_finger4'), id=15, color=[255, 51, 121 | 51]), 122 | 16: 123 | dict(link=('wrist', 'pinky_finger1'), id=16, color=[0, 255, 0]), 124 | 17: 125 | dict(link=('pinky_finger1', 'pinky_finger2'), id=17, color=[0, 255, 126 | 0]), 127 | 18: 128 | dict(link=('pinky_finger2', 'pinky_finger3'), id=18, color=[0, 255, 129 | 0]), 130 | 19: 131 | dict(link=('pinky_finger3', 'pinky_finger4'), id=19, color=[0, 255, 0]) 132 | }, 133 | joint_weights=[1.] * 21, 134 | sigmas=[]) 135 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/onehand10k.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='onehand10k', 3 | paper_info=dict( 4 | author='Wang, Yangang and Peng, Cong and Liu, Yebin', 5 | title='Mask-pose cascaded cnn for 2d hand pose estimation ' 6 | 'from single color image', 7 | container='IEEE Transactions on Circuits and Systems ' 8 | 'for Video Technology', 9 | year='2018', 10 | homepage='https://www.yangangwang.com/papers/WANG-MCC-2018-10.html', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='wrist', id=0, color=[255, 255, 255], type='', swap=''), 15 | 1: 16 | dict(name='thumb1', id=1, color=[255, 128, 0], type='', swap=''), 17 | 2: 18 | dict(name='thumb2', id=2, color=[255, 128, 0], type='', swap=''), 19 | 3: 20 | dict(name='thumb3', id=3, color=[255, 128, 0], type='', swap=''), 21 | 4: 22 | dict(name='thumb4', id=4, color=[255, 128, 0], type='', swap=''), 23 | 5: 24 | dict(name='forefinger1', id=5, color=[255, 153, 255], type='', 25 | swap=''), 26 | 6: 27 | dict(name='forefinger2', id=6, color=[255, 153, 255], type='', 28 | swap=''), 29 | 7: 30 | dict(name='forefinger3', id=7, color=[255, 153, 255], type='', 31 | swap=''), 32 | 8: 33 | dict(name='forefinger4', id=8, color=[255, 153, 255], type='', 34 | swap=''), 35 | 9: 36 | dict(name='middle_finger1', 37 | id=9, 38 | color=[102, 178, 255], 39 | type='', 40 | swap=''), 41 | 10: 42 | dict(name='middle_finger2', 43 | id=10, 44 | color=[102, 178, 255], 45 | type='', 46 | swap=''), 47 | 11: 48 | dict(name='middle_finger3', 49 | id=11, 50 | color=[102, 178, 255], 51 | type='', 52 | swap=''), 53 | 12: 54 | dict(name='middle_finger4', 55 | id=12, 56 | color=[102, 178, 255], 57 | type='', 58 | swap=''), 59 | 13: 60 | dict(name='ring_finger1', id=13, color=[255, 51, 51], type='', 61 | swap=''), 62 | 14: 63 | dict(name='ring_finger2', id=14, color=[255, 51, 51], type='', 64 | swap=''), 65 | 15: 66 | dict(name='ring_finger3', id=15, color=[255, 51, 51], type='', 67 | swap=''), 68 | 16: 69 | dict(name='ring_finger4', id=16, color=[255, 51, 51], type='', 70 | swap=''), 71 | 17: 72 | dict(name='pinky_finger1', id=17, color=[0, 255, 0], type='', swap=''), 73 | 18: 74 | dict(name='pinky_finger2', id=18, color=[0, 255, 0], type='', swap=''), 75 | 19: 76 | dict(name='pinky_finger3', id=19, color=[0, 255, 0], type='', swap=''), 77 | 20: 78 | dict(name='pinky_finger4', id=20, color=[0, 255, 0], type='', swap='') 79 | }, 80 | skeleton_info={ 81 | 0: 82 | dict(link=('wrist', 'thumb1'), id=0, color=[255, 128, 0]), 83 | 1: 84 | dict(link=('thumb1', 'thumb2'), id=1, color=[255, 128, 0]), 85 | 2: 86 | dict(link=('thumb2', 'thumb3'), id=2, color=[255, 128, 0]), 87 | 3: 88 | dict(link=('thumb3', 'thumb4'), id=3, color=[255, 128, 0]), 89 | 4: 90 | dict(link=('wrist', 'forefinger1'), id=4, color=[255, 153, 255]), 91 | 5: 92 | dict(link=('forefinger1', 'forefinger2'), id=5, color=[255, 153, 255]), 93 | 6: 94 | dict(link=('forefinger2', 'forefinger3'), id=6, color=[255, 153, 255]), 95 | 7: 96 | dict(link=('forefinger3', 'forefinger4'), id=7, color=[255, 153, 255]), 97 | 8: 98 | dict(link=('wrist', 'middle_finger1'), id=8, color=[102, 178, 255]), 99 | 9: 100 | dict(link=('middle_finger1', 'middle_finger2'), 101 | id=9, 102 | color=[102, 178, 255]), 103 | 10: 104 | dict(link=('middle_finger2', 'middle_finger3'), 105 | id=10, 106 | color=[102, 178, 255]), 107 | 11: 108 | dict(link=('middle_finger3', 'middle_finger4'), 109 | id=11, 110 | color=[102, 178, 255]), 111 | 12: 112 | dict(link=('wrist', 'ring_finger1'), id=12, color=[255, 51, 51]), 113 | 13: 114 | dict(link=('ring_finger1', 'ring_finger2'), id=13, color=[255, 51, 115 | 51]), 116 | 14: 117 | dict(link=('ring_finger2', 'ring_finger3'), id=14, color=[255, 51, 118 | 51]), 119 | 15: 120 | dict(link=('ring_finger3', 'ring_finger4'), id=15, color=[255, 51, 121 | 51]), 122 | 16: 123 | dict(link=('wrist', 'pinky_finger1'), id=16, color=[0, 255, 0]), 124 | 17: 125 | dict(link=('pinky_finger1', 'pinky_finger2'), id=17, color=[0, 255, 126 | 0]), 127 | 18: 128 | dict(link=('pinky_finger2', 'pinky_finger3'), id=18, color=[0, 255, 129 | 0]), 130 | 19: 131 | dict(link=('pinky_finger3', 'pinky_finger4'), id=19, color=[0, 255, 0]) 132 | }, 133 | joint_weights=[1.] * 21, 134 | sigmas=[]) 135 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/panoptic_body3d.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='panoptic_pose_3d', 3 | paper_info=dict( 4 | author='Joo, Hanbyul and Simon, Tomas and Li, Xulong' 5 | 'and Liu, Hao and Tan, Lei and Gui, Lin and Banerjee, Sean' 6 | 'and Godisart, Timothy and Nabbe, Bart and Matthews, Iain' 7 | 'and Kanade, Takeo and Nobuhara, Shohei and Sheikh, Yaser', 8 | title='Panoptic Studio: A Massively Multiview System ' 9 | 'for Interaction Motion Capture', 10 | container='IEEE Transactions on Pattern Analysis' 11 | ' and Machine Intelligence', 12 | year='2017', 13 | homepage='http://domedb.perception.cs.cmu.edu', 14 | ), 15 | keypoint_info={ 16 | 0: 17 | dict(name='neck', id=0, color=[51, 153, 255], type='upper', swap=''), 18 | 1: 19 | dict(name='nose', id=1, color=[51, 153, 255], type='upper', swap=''), 20 | 2: 21 | dict(name='mid_hip', id=2, color=[0, 255, 0], type='lower', swap=''), 22 | 3: 23 | dict(name='left_shoulder', 24 | id=3, 25 | color=[0, 255, 0], 26 | type='upper', 27 | swap='right_shoulder'), 28 | 4: 29 | dict(name='left_elbow', 30 | id=4, 31 | color=[0, 255, 0], 32 | type='upper', 33 | swap='right_elbow'), 34 | 5: 35 | dict(name='left_wrist', 36 | id=5, 37 | color=[0, 255, 0], 38 | type='upper', 39 | swap='right_wrist'), 40 | 6: 41 | dict(name='left_hip', 42 | id=6, 43 | color=[0, 255, 0], 44 | type='lower', 45 | swap='right_hip'), 46 | 7: 47 | dict(name='left_knee', 48 | id=7, 49 | color=[0, 255, 0], 50 | type='lower', 51 | swap='right_knee'), 52 | 8: 53 | dict(name='left_ankle', 54 | id=8, 55 | color=[0, 255, 0], 56 | type='lower', 57 | swap='right_ankle'), 58 | 9: 59 | dict(name='right_shoulder', 60 | id=9, 61 | color=[255, 128, 0], 62 | type='upper', 63 | swap='left_shoulder'), 64 | 10: 65 | dict(name='right_elbow', 66 | id=10, 67 | color=[255, 128, 0], 68 | type='upper', 69 | swap='left_elbow'), 70 | 11: 71 | dict(name='right_wrist', 72 | id=11, 73 | color=[255, 128, 0], 74 | type='upper', 75 | swap='left_wrist'), 76 | 12: 77 | dict(name='right_hip', 78 | id=12, 79 | color=[255, 128, 0], 80 | type='lower', 81 | swap='left_hip'), 82 | 13: 83 | dict(name='right_knee', 84 | id=13, 85 | color=[255, 128, 0], 86 | type='lower', 87 | swap='left_knee'), 88 | 14: 89 | dict(name='right_ankle', 90 | id=14, 91 | color=[255, 128, 0], 92 | type='lower', 93 | swap='left_ankle'), 94 | 15: 95 | dict(name='left_eye', 96 | id=15, 97 | color=[51, 153, 255], 98 | type='upper', 99 | swap='right_eye'), 100 | 16: 101 | dict(name='left_ear', 102 | id=16, 103 | color=[51, 153, 255], 104 | type='upper', 105 | swap='right_ear'), 106 | 17: 107 | dict(name='right_eye', 108 | id=17, 109 | color=[51, 153, 255], 110 | type='upper', 111 | swap='left_eye'), 112 | 18: 113 | dict(name='right_ear', 114 | id=18, 115 | color=[51, 153, 255], 116 | type='upper', 117 | swap='left_ear') 118 | }, 119 | skeleton_info={ 120 | 0: dict(link=('nose', 'neck'), id=0, color=[51, 153, 255]), 121 | 1: dict(link=('neck', 'left_shoulder'), id=1, color=[0, 255, 0]), 122 | 2: dict(link=('neck', 'right_shoulder'), id=2, color=[255, 128, 0]), 123 | 3: dict(link=('left_shoulder', 'left_elbow'), id=3, color=[0, 255, 0]), 124 | 4: dict(link=('right_shoulder', 'right_elbow'), 125 | id=4, 126 | color=[255, 128, 0]), 127 | 5: dict(link=('left_elbow', 'left_wrist'), id=5, color=[0, 255, 0]), 128 | 6: dict(link=('right_elbow', 'right_wrist'), id=6, color=[255, 128, 129 | 0]), 130 | 7: dict(link=('left_ankle', 'left_knee'), id=7, color=[0, 255, 0]), 131 | 8: dict(link=('left_knee', 'left_hip'), id=8, color=[0, 255, 0]), 132 | 9: dict(link=('right_ankle', 'right_knee'), id=9, color=[255, 128, 0]), 133 | 10: dict(link=('right_knee', 'right_hip'), id=10, color=[255, 128, 0]), 134 | 11: dict(link=('mid_hip', 'left_hip'), id=11, color=[0, 255, 0]), 135 | 12: dict(link=('mid_hip', 'right_hip'), id=12, color=[255, 128, 0]), 136 | 13: dict(link=('mid_hip', 'neck'), id=13, color=[51, 153, 255]), 137 | }, 138 | joint_weights=[ 139 | 1.0, 1.0, 1.0, 1.0, 1.2, 1.5, 1.0, 1.2, 1.5, 1.0, 1.2, 1.5, 1.0, 1.2, 140 | 1.5, 1.0, 1.0, 1.0, 1.0 141 | ], 142 | sigmas=[ 143 | 0.026, 0.026, 0.107, 0.079, 0.072, 0.062, 0.107, 0.087, 0.089, 0.079, 144 | 0.072, 0.062, 0.107, 0.087, 0.089, 0.025, 0.035, 0.025, 0.035 145 | ]) 146 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/panoptic_hand2d.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='panoptic_hand2d', 3 | paper_info=dict( 4 | author='Simon, Tomas and Joo, Hanbyul and ' 5 | 'Matthews, Iain and Sheikh, Yaser', 6 | title='Hand keypoint detection in single images using ' 7 | 'multiview bootstrapping', 8 | container='Proceedings of the IEEE conference on ' 9 | 'Computer Vision and Pattern Recognition', 10 | year='2017', 11 | homepage='http://domedb.perception.cs.cmu.edu/handdb.html', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='wrist', id=0, color=[255, 255, 255], type='', swap=''), 16 | 1: 17 | dict(name='thumb1', id=1, color=[255, 128, 0], type='', swap=''), 18 | 2: 19 | dict(name='thumb2', id=2, color=[255, 128, 0], type='', swap=''), 20 | 3: 21 | dict(name='thumb3', id=3, color=[255, 128, 0], type='', swap=''), 22 | 4: 23 | dict(name='thumb4', id=4, color=[255, 128, 0], type='', swap=''), 24 | 5: 25 | dict(name='forefinger1', id=5, color=[255, 153, 255], type='', 26 | swap=''), 27 | 6: 28 | dict(name='forefinger2', id=6, color=[255, 153, 255], type='', 29 | swap=''), 30 | 7: 31 | dict(name='forefinger3', id=7, color=[255, 153, 255], type='', 32 | swap=''), 33 | 8: 34 | dict(name='forefinger4', id=8, color=[255, 153, 255], type='', 35 | swap=''), 36 | 9: 37 | dict(name='middle_finger1', 38 | id=9, 39 | color=[102, 178, 255], 40 | type='', 41 | swap=''), 42 | 10: 43 | dict(name='middle_finger2', 44 | id=10, 45 | color=[102, 178, 255], 46 | type='', 47 | swap=''), 48 | 11: 49 | dict(name='middle_finger3', 50 | id=11, 51 | color=[102, 178, 255], 52 | type='', 53 | swap=''), 54 | 12: 55 | dict(name='middle_finger4', 56 | id=12, 57 | color=[102, 178, 255], 58 | type='', 59 | swap=''), 60 | 13: 61 | dict(name='ring_finger1', id=13, color=[255, 51, 51], type='', 62 | swap=''), 63 | 14: 64 | dict(name='ring_finger2', id=14, color=[255, 51, 51], type='', 65 | swap=''), 66 | 15: 67 | dict(name='ring_finger3', id=15, color=[255, 51, 51], type='', 68 | swap=''), 69 | 16: 70 | dict(name='ring_finger4', id=16, color=[255, 51, 51], type='', 71 | swap=''), 72 | 17: 73 | dict(name='pinky_finger1', id=17, color=[0, 255, 0], type='', swap=''), 74 | 18: 75 | dict(name='pinky_finger2', id=18, color=[0, 255, 0], type='', swap=''), 76 | 19: 77 | dict(name='pinky_finger3', id=19, color=[0, 255, 0], type='', swap=''), 78 | 20: 79 | dict(name='pinky_finger4', id=20, color=[0, 255, 0], type='', swap='') 80 | }, 81 | skeleton_info={ 82 | 0: 83 | dict(link=('wrist', 'thumb1'), id=0, color=[255, 128, 0]), 84 | 1: 85 | dict(link=('thumb1', 'thumb2'), id=1, color=[255, 128, 0]), 86 | 2: 87 | dict(link=('thumb2', 'thumb3'), id=2, color=[255, 128, 0]), 88 | 3: 89 | dict(link=('thumb3', 'thumb4'), id=3, color=[255, 128, 0]), 90 | 4: 91 | dict(link=('wrist', 'forefinger1'), id=4, color=[255, 153, 255]), 92 | 5: 93 | dict(link=('forefinger1', 'forefinger2'), id=5, color=[255, 153, 255]), 94 | 6: 95 | dict(link=('forefinger2', 'forefinger3'), id=6, color=[255, 153, 255]), 96 | 7: 97 | dict(link=('forefinger3', 'forefinger4'), id=7, color=[255, 153, 255]), 98 | 8: 99 | dict(link=('wrist', 'middle_finger1'), id=8, color=[102, 178, 255]), 100 | 9: 101 | dict(link=('middle_finger1', 'middle_finger2'), 102 | id=9, 103 | color=[102, 178, 255]), 104 | 10: 105 | dict(link=('middle_finger2', 'middle_finger3'), 106 | id=10, 107 | color=[102, 178, 255]), 108 | 11: 109 | dict(link=('middle_finger3', 'middle_finger4'), 110 | id=11, 111 | color=[102, 178, 255]), 112 | 12: 113 | dict(link=('wrist', 'ring_finger1'), id=12, color=[255, 51, 51]), 114 | 13: 115 | dict(link=('ring_finger1', 'ring_finger2'), id=13, color=[255, 51, 116 | 51]), 117 | 14: 118 | dict(link=('ring_finger2', 'ring_finger3'), id=14, color=[255, 51, 119 | 51]), 120 | 15: 121 | dict(link=('ring_finger3', 'ring_finger4'), id=15, color=[255, 51, 122 | 51]), 123 | 16: 124 | dict(link=('wrist', 'pinky_finger1'), id=16, color=[0, 255, 0]), 125 | 17: 126 | dict(link=('pinky_finger1', 'pinky_finger2'), id=17, color=[0, 255, 127 | 0]), 128 | 18: 129 | dict(link=('pinky_finger2', 'pinky_finger3'), id=18, color=[0, 255, 130 | 0]), 131 | 19: 132 | dict(link=('pinky_finger3', 'pinky_finger4'), id=19, color=[0, 255, 0]) 133 | }, 134 | joint_weights=[1.] * 21, 135 | sigmas=[]) 136 | -------------------------------------------------------------------------------- /configs/examples/pose_estimation/webcam_demo.md: -------------------------------------------------------------------------------- 1 | ## Webcam Demo 2 | 3 | This is a webcam demo which integrartes detection and 2D pose estimation for humans and animals. It can also apply fun effects like putting on sunglasses or enlarging the eyes, based on the pose estimation results. 4 | 5 |
6 |
7 |
8 | 9 | ### Get started 10 | 11 | Launch the demo: 12 | 13 | ```shell 14 | python run.py --config configs/examples/pose_estimation/pose_estimation.py 15 | ``` 16 | 17 | The command above will use the config file `configs/examples/pose_estimation/pose_estimation.py`. You can also specify the config file in the command: 18 | 19 | ```shell 20 | # Use the config "pose_tracking.py" for higher infererence speed 21 | python run.py --config configs/examples/pose_estimation/pose_tracking.py 22 | ``` 23 | 24 | ### Hotkeys 25 | 26 | | Hotkey | Function | 27 | | ------ | ---------------------------------------------------------------- | 28 | | v | Toggle the pose visualization on/off. | 29 | | s | Toggle the sunglasses effect on/off. (NA for `pose_trakcing.py`) | 30 | | b | Toggle the bug-eye effect on/off. (NA for `pose_trakcing.py`) | 31 | | h | Show help information. | 32 | | m | Show the monitoring information. | 33 | | q | Exit. | 34 | 35 | Note that the demo will automatically save the output video into a file `webcam_demo.mp4`. 36 | 37 | ### Usage and configuarations 38 | 39 | Detailed configurations can be found in the config file. 40 | 41 | - **Configure detection models** 42 | Users can choose detection models from the [MMDetection Model Zoo](https://mmdetection.readthedocs.io/en/v2.20.0/model_zoo.html). Just set the `model_config` and `model_checkpoint` in the detector node accordingly, and the model will be automatically downloaded and loaded. 43 | 44 | ```python 45 | # 'DetectorNode': 46 | # This node performs object detection from the frame image using an 47 | # MMDetection model. 48 | dict( 49 | type='DetectorNode', 50 | name='detector', 51 | model_config='model_configs/mmdet/' 52 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 53 | model_checkpoint='https://download.openmmlab.com' 54 | '/mmdetection/v2.0/ssd/' 55 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 56 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 57 | input_buffer='_input_', 58 | output_buffer='det_result') 59 | ``` 60 | 61 | - **Configure pose estimation models** 62 | In this demo we use two [top-down](https://github.com/open-mmlab/mmpose/tree/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap) pose estimation models for humans and animals respectively. Users can choose models from the [MMPose Model Zoo](https://mmpose.readthedocs.io/en/latest/modelzoo.html). To apply different pose models on different instance types, you can add multiple pose estimator nodes with `cls_names` set accordingly. 63 | 64 | ```python 65 | # 'TopDownPoseEstimatorNode': 66 | # This node performs keypoint detection from the frame image using an 67 | # MMPose top-down model. Detection results is needed. 68 | dict( 69 | type='TopDownPoseEstimatorNode', 70 | name='human pose estimator', 71 | model_config='configs/wholebody/2d_kpt_sview_rgb_img/' 72 | 'topdown_heatmap/coco-wholebody/' 73 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 74 | model_checkpoint='model_configs/mmpose/vipnas_mbv3_co' 75 | 'co_wholebody_256x192_dark-e2158108_20211205.pth', 76 | labels=['person'], 77 | input_buffer='det_result', 78 | output_buffer='human_pose'), 79 | dict( 80 | type='TopDownPoseEstimatorNode', 81 | name='animal pose estimator', 82 | model_config='model_configs/mmdet/hrnet_w32_animalpose_256x256.py', 83 | model_checkpoint='https://download.openmmlab.com/mmpose/animal/' 84 | 'hrnet/hrnet_w32_animalpose_256x256-1aa7f075_20210426.pth', 85 | labels=['cat', 'dog', 'horse', 'sheep', 'cow'], 86 | input_buffer='human_pose', 87 | output_buffer='animal_pose') 88 | ``` 89 | 90 | - **Run the demo without GPU** 91 | If you don't have GPU and CUDA in your device, the demo can run with only CPU by setting `device='cpu'` in all model nodes. For example: 92 | 93 | ```python 94 | dict( 95 | type='DetectorNode', 96 | name='detector', 97 | model_config='model_configs/mmdet/' 98 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 99 | model_checkpoint='https://download.openmmlab.com' 100 | '/mmdetection/v2.0/ssd/' 101 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 102 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 103 | device='cpu', 104 | input_buffer='_input_', 105 | output_buffer='det_result') 106 | ``` 107 | 108 | - **Run the demo on a local video file** 109 | You can use local video files as the demo input by set `camera_id` to the file path. 110 | 111 | - **The computer doesn't have a camera?** 112 | A smart phone can serve as a webcam via apps like [Camo](https://reincubate.com/camo/) or [DroidCam](https://www.dev47apps.com/). 113 | 114 | - **Test the camera and display** 115 | Run follow command for a quick test of video capturing and displaying. 116 | 117 | ```shell 118 | python run.py --config configs/examples/pose_estimation/test_camera.py 119 | ``` 120 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/freihand2d.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='freihand', 3 | paper_info=dict( 4 | author='Zimmermann, Christian and Ceylan, Duygu and ' 5 | 'Yang, Jimei and Russell, Bryan and ' 6 | 'Argus, Max and Brox, Thomas', 7 | title='Freihand: A dataset for markerless capture of hand pose ' 8 | 'and shape from single rgb images', 9 | container='Proceedings of the IEEE International ' 10 | 'Conference on Computer Vision', 11 | year='2019', 12 | homepage='https://lmb.informatik.uni-freiburg.de/projects/freihand/', 13 | ), 14 | keypoint_info={ 15 | 0: 16 | dict(name='wrist', id=0, color=[255, 255, 255], type='', swap=''), 17 | 1: 18 | dict(name='thumb1', id=1, color=[255, 128, 0], type='', swap=''), 19 | 2: 20 | dict(name='thumb2', id=2, color=[255, 128, 0], type='', swap=''), 21 | 3: 22 | dict(name='thumb3', id=3, color=[255, 128, 0], type='', swap=''), 23 | 4: 24 | dict(name='thumb4', id=4, color=[255, 128, 0], type='', swap=''), 25 | 5: 26 | dict(name='forefinger1', id=5, color=[255, 153, 255], type='', 27 | swap=''), 28 | 6: 29 | dict(name='forefinger2', id=6, color=[255, 153, 255], type='', 30 | swap=''), 31 | 7: 32 | dict(name='forefinger3', id=7, color=[255, 153, 255], type='', 33 | swap=''), 34 | 8: 35 | dict(name='forefinger4', id=8, color=[255, 153, 255], type='', 36 | swap=''), 37 | 9: 38 | dict(name='middle_finger1', 39 | id=9, 40 | color=[102, 178, 255], 41 | type='', 42 | swap=''), 43 | 10: 44 | dict(name='middle_finger2', 45 | id=10, 46 | color=[102, 178, 255], 47 | type='', 48 | swap=''), 49 | 11: 50 | dict(name='middle_finger3', 51 | id=11, 52 | color=[102, 178, 255], 53 | type='', 54 | swap=''), 55 | 12: 56 | dict(name='middle_finger4', 57 | id=12, 58 | color=[102, 178, 255], 59 | type='', 60 | swap=''), 61 | 13: 62 | dict(name='ring_finger1', id=13, color=[255, 51, 51], type='', 63 | swap=''), 64 | 14: 65 | dict(name='ring_finger2', id=14, color=[255, 51, 51], type='', 66 | swap=''), 67 | 15: 68 | dict(name='ring_finger3', id=15, color=[255, 51, 51], type='', 69 | swap=''), 70 | 16: 71 | dict(name='ring_finger4', id=16, color=[255, 51, 51], type='', 72 | swap=''), 73 | 17: 74 | dict(name='pinky_finger1', id=17, color=[0, 255, 0], type='', swap=''), 75 | 18: 76 | dict(name='pinky_finger2', id=18, color=[0, 255, 0], type='', swap=''), 77 | 19: 78 | dict(name='pinky_finger3', id=19, color=[0, 255, 0], type='', swap=''), 79 | 20: 80 | dict(name='pinky_finger4', id=20, color=[0, 255, 0], type='', swap='') 81 | }, 82 | skeleton_info={ 83 | 0: 84 | dict(link=('wrist', 'thumb1'), id=0, color=[255, 128, 0]), 85 | 1: 86 | dict(link=('thumb1', 'thumb2'), id=1, color=[255, 128, 0]), 87 | 2: 88 | dict(link=('thumb2', 'thumb3'), id=2, color=[255, 128, 0]), 89 | 3: 90 | dict(link=('thumb3', 'thumb4'), id=3, color=[255, 128, 0]), 91 | 4: 92 | dict(link=('wrist', 'forefinger1'), id=4, color=[255, 153, 255]), 93 | 5: 94 | dict(link=('forefinger1', 'forefinger2'), id=5, color=[255, 153, 255]), 95 | 6: 96 | dict(link=('forefinger2', 'forefinger3'), id=6, color=[255, 153, 255]), 97 | 7: 98 | dict(link=('forefinger3', 'forefinger4'), id=7, color=[255, 153, 255]), 99 | 8: 100 | dict(link=('wrist', 'middle_finger1'), id=8, color=[102, 178, 255]), 101 | 9: 102 | dict(link=('middle_finger1', 'middle_finger2'), 103 | id=9, 104 | color=[102, 178, 255]), 105 | 10: 106 | dict(link=('middle_finger2', 'middle_finger3'), 107 | id=10, 108 | color=[102, 178, 255]), 109 | 11: 110 | dict(link=('middle_finger3', 'middle_finger4'), 111 | id=11, 112 | color=[102, 178, 255]), 113 | 12: 114 | dict(link=('wrist', 'ring_finger1'), id=12, color=[255, 51, 51]), 115 | 13: 116 | dict(link=('ring_finger1', 'ring_finger2'), id=13, color=[255, 51, 117 | 51]), 118 | 14: 119 | dict(link=('ring_finger2', 'ring_finger3'), id=14, color=[255, 51, 120 | 51]), 121 | 15: 122 | dict(link=('ring_finger3', 'ring_finger4'), id=15, color=[255, 51, 123 | 51]), 124 | 16: 125 | dict(link=('wrist', 'pinky_finger1'), id=16, color=[0, 255, 0]), 126 | 17: 127 | dict(link=('pinky_finger1', 'pinky_finger2'), id=17, color=[0, 255, 128 | 0]), 129 | 18: 130 | dict(link=('pinky_finger2', 'pinky_finger3'), id=18, color=[0, 255, 131 | 0]), 132 | 19: 133 | dict(link=('pinky_finger3', 'pinky_finger4'), id=19, color=[0, 255, 0]) 134 | }, 135 | joint_weights=[1.] * 21, 136 | sigmas=[]) 137 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/posetrack18.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='posetrack18', 3 | paper_info=dict( 4 | author='Andriluka, Mykhaylo and Iqbal, Umar and ' 5 | 'Insafutdinov, Eldar and Pishchulin, Leonid and ' 6 | 'Milan, Anton and Gall, Juergen and Schiele, Bernt', 7 | title='Posetrack: A benchmark for human pose estimation and tracking', 8 | container='Proceedings of the IEEE Conference on ' 9 | 'Computer Vision and Pattern Recognition', 10 | year='2018', 11 | homepage='https://posetrack.net/users/download.php', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='nose', id=0, color=[51, 153, 255], type='upper', swap=''), 16 | 1: 17 | dict(name='head_bottom', 18 | id=1, 19 | color=[51, 153, 255], 20 | type='upper', 21 | swap=''), 22 | 2: 23 | dict(name='head_top', 24 | id=2, 25 | color=[51, 153, 255], 26 | type='upper', 27 | swap=''), 28 | 3: 29 | dict(name='left_ear', 30 | id=3, 31 | color=[51, 153, 255], 32 | type='upper', 33 | swap='right_ear'), 34 | 4: 35 | dict(name='right_ear', 36 | id=4, 37 | color=[51, 153, 255], 38 | type='upper', 39 | swap='left_ear'), 40 | 5: 41 | dict(name='left_shoulder', 42 | id=5, 43 | color=[0, 255, 0], 44 | type='upper', 45 | swap='right_shoulder'), 46 | 6: 47 | dict(name='right_shoulder', 48 | id=6, 49 | color=[255, 128, 0], 50 | type='upper', 51 | swap='left_shoulder'), 52 | 7: 53 | dict(name='left_elbow', 54 | id=7, 55 | color=[0, 255, 0], 56 | type='upper', 57 | swap='right_elbow'), 58 | 8: 59 | dict(name='right_elbow', 60 | id=8, 61 | color=[255, 128, 0], 62 | type='upper', 63 | swap='left_elbow'), 64 | 9: 65 | dict(name='left_wrist', 66 | id=9, 67 | color=[0, 255, 0], 68 | type='upper', 69 | swap='right_wrist'), 70 | 10: 71 | dict(name='right_wrist', 72 | id=10, 73 | color=[255, 128, 0], 74 | type='upper', 75 | swap='left_wrist'), 76 | 11: 77 | dict(name='left_hip', 78 | id=11, 79 | color=[0, 255, 0], 80 | type='lower', 81 | swap='right_hip'), 82 | 12: 83 | dict(name='right_hip', 84 | id=12, 85 | color=[255, 128, 0], 86 | type='lower', 87 | swap='left_hip'), 88 | 13: 89 | dict(name='left_knee', 90 | id=13, 91 | color=[0, 255, 0], 92 | type='lower', 93 | swap='right_knee'), 94 | 14: 95 | dict(name='right_knee', 96 | id=14, 97 | color=[255, 128, 0], 98 | type='lower', 99 | swap='left_knee'), 100 | 15: 101 | dict(name='left_ankle', 102 | id=15, 103 | color=[0, 255, 0], 104 | type='lower', 105 | swap='right_ankle'), 106 | 16: 107 | dict(name='right_ankle', 108 | id=16, 109 | color=[255, 128, 0], 110 | type='lower', 111 | swap='left_ankle') 112 | }, 113 | skeleton_info={ 114 | 0: 115 | dict(link=('left_ankle', 'left_knee'), id=0, color=[0, 255, 0]), 116 | 1: 117 | dict(link=('left_knee', 'left_hip'), id=1, color=[0, 255, 0]), 118 | 2: 119 | dict(link=('right_ankle', 'right_knee'), id=2, color=[255, 128, 0]), 120 | 3: 121 | dict(link=('right_knee', 'right_hip'), id=3, color=[255, 128, 0]), 122 | 4: 123 | dict(link=('left_hip', 'right_hip'), id=4, color=[51, 153, 255]), 124 | 5: 125 | dict(link=('left_shoulder', 'left_hip'), id=5, color=[51, 153, 255]), 126 | 6: 127 | dict(link=('right_shoulder', 'right_hip'), id=6, color=[51, 153, 255]), 128 | 7: 129 | dict(link=('left_shoulder', 'right_shoulder'), 130 | id=7, 131 | color=[51, 153, 255]), 132 | 8: 133 | dict(link=('left_shoulder', 'left_elbow'), id=8, color=[0, 255, 0]), 134 | 9: 135 | dict(link=('right_shoulder', 'right_elbow'), id=9, color=[255, 128, 136 | 0]), 137 | 10: 138 | dict(link=('left_elbow', 'left_wrist'), id=10, color=[0, 255, 0]), 139 | 11: 140 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 141 | 12: 142 | dict(link=('nose', 'head_bottom'), id=12, color=[51, 153, 255]), 143 | 13: 144 | dict(link=('nose', 'head_top'), id=13, color=[51, 153, 255]), 145 | 14: 146 | dict(link=('head_bottom', 'left_shoulder'), 147 | id=14, 148 | color=[51, 153, 255]), 149 | 15: 150 | dict(link=('head_bottom', 'right_shoulder'), 151 | id=15, 152 | color=[51, 153, 255]) 153 | }, 154 | joint_weights=[ 155 | 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.5, 1.5, 1., 1., 1.2, 1.2, 1.5, 156 | 1.5 157 | ], 158 | sigmas=[ 159 | 0.026, 0.025, 0.025, 0.035, 0.035, 0.079, 0.079, 0.072, 0.072, 0.062, 160 | 0.062, 0.107, 0.107, 0.087, 0.087, 0.089, 0.089 161 | ]) 162 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/ochuman.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='ochuman', 3 | paper_info=dict( 4 | author='Zhang, Song-Hai and Li, Ruilong and Dong, Xin and ' 5 | 'Rosin, Paul and Cai, Zixi and Han, Xi and ' 6 | 'Yang, Dingcheng and Huang, Haozhi and Hu, Shi-Min', 7 | title='Pose2seg: Detection free human instance segmentation', 8 | container='Proceedings of the IEEE conference on computer ' 9 | 'vision and pattern recognition', 10 | year='2019', 11 | homepage='https://github.com/liruilong940607/OCHumanApi', 12 | ), 13 | keypoint_info={ 14 | 0: 15 | dict(name='nose', id=0, color=[51, 153, 255], type='upper', swap=''), 16 | 1: 17 | dict(name='left_eye', 18 | id=1, 19 | color=[51, 153, 255], 20 | type='upper', 21 | swap='right_eye'), 22 | 2: 23 | dict(name='right_eye', 24 | id=2, 25 | color=[51, 153, 255], 26 | type='upper', 27 | swap='left_eye'), 28 | 3: 29 | dict(name='left_ear', 30 | id=3, 31 | color=[51, 153, 255], 32 | type='upper', 33 | swap='right_ear'), 34 | 4: 35 | dict(name='right_ear', 36 | id=4, 37 | color=[51, 153, 255], 38 | type='upper', 39 | swap='left_ear'), 40 | 5: 41 | dict(name='left_shoulder', 42 | id=5, 43 | color=[0, 255, 0], 44 | type='upper', 45 | swap='right_shoulder'), 46 | 6: 47 | dict(name='right_shoulder', 48 | id=6, 49 | color=[255, 128, 0], 50 | type='upper', 51 | swap='left_shoulder'), 52 | 7: 53 | dict(name='left_elbow', 54 | id=7, 55 | color=[0, 255, 0], 56 | type='upper', 57 | swap='right_elbow'), 58 | 8: 59 | dict(name='right_elbow', 60 | id=8, 61 | color=[255, 128, 0], 62 | type='upper', 63 | swap='left_elbow'), 64 | 9: 65 | dict(name='left_wrist', 66 | id=9, 67 | color=[0, 255, 0], 68 | type='upper', 69 | swap='right_wrist'), 70 | 10: 71 | dict(name='right_wrist', 72 | id=10, 73 | color=[255, 128, 0], 74 | type='upper', 75 | swap='left_wrist'), 76 | 11: 77 | dict(name='left_hip', 78 | id=11, 79 | color=[0, 255, 0], 80 | type='lower', 81 | swap='right_hip'), 82 | 12: 83 | dict(name='right_hip', 84 | id=12, 85 | color=[255, 128, 0], 86 | type='lower', 87 | swap='left_hip'), 88 | 13: 89 | dict(name='left_knee', 90 | id=13, 91 | color=[0, 255, 0], 92 | type='lower', 93 | swap='right_knee'), 94 | 14: 95 | dict(name='right_knee', 96 | id=14, 97 | color=[255, 128, 0], 98 | type='lower', 99 | swap='left_knee'), 100 | 15: 101 | dict(name='left_ankle', 102 | id=15, 103 | color=[0, 255, 0], 104 | type='lower', 105 | swap='right_ankle'), 106 | 16: 107 | dict(name='right_ankle', 108 | id=16, 109 | color=[255, 128, 0], 110 | type='lower', 111 | swap='left_ankle') 112 | }, 113 | skeleton_info={ 114 | 0: 115 | dict(link=('left_ankle', 'left_knee'), id=0, color=[0, 255, 0]), 116 | 1: 117 | dict(link=('left_knee', 'left_hip'), id=1, color=[0, 255, 0]), 118 | 2: 119 | dict(link=('right_ankle', 'right_knee'), id=2, color=[255, 128, 0]), 120 | 3: 121 | dict(link=('right_knee', 'right_hip'), id=3, color=[255, 128, 0]), 122 | 4: 123 | dict(link=('left_hip', 'right_hip'), id=4, color=[51, 153, 255]), 124 | 5: 125 | dict(link=('left_shoulder', 'left_hip'), id=5, color=[51, 153, 255]), 126 | 6: 127 | dict(link=('right_shoulder', 'right_hip'), id=6, color=[51, 153, 255]), 128 | 7: 129 | dict(link=('left_shoulder', 'right_shoulder'), 130 | id=7, 131 | color=[51, 153, 255]), 132 | 8: 133 | dict(link=('left_shoulder', 'left_elbow'), id=8, color=[0, 255, 0]), 134 | 9: 135 | dict(link=('right_shoulder', 'right_elbow'), id=9, color=[255, 128, 136 | 0]), 137 | 10: 138 | dict(link=('left_elbow', 'left_wrist'), id=10, color=[0, 255, 0]), 139 | 11: 140 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 141 | 12: 142 | dict(link=('left_eye', 'right_eye'), id=12, color=[51, 153, 255]), 143 | 13: 144 | dict(link=('nose', 'left_eye'), id=13, color=[51, 153, 255]), 145 | 14: 146 | dict(link=('nose', 'right_eye'), id=14, color=[51, 153, 255]), 147 | 15: 148 | dict(link=('left_eye', 'left_ear'), id=15, color=[51, 153, 255]), 149 | 16: 150 | dict(link=('right_eye', 'right_ear'), id=16, color=[51, 153, 255]), 151 | 17: 152 | dict(link=('left_ear', 'left_shoulder'), id=17, color=[51, 153, 255]), 153 | 18: 154 | dict(link=('right_ear', 'right_shoulder'), id=18, color=[51, 153, 255]) 155 | }, 156 | joint_weights=[ 157 | 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.5, 1.5, 1., 1., 1.2, 1.2, 1.5, 158 | 1.5 159 | ], 160 | sigmas=[ 161 | 0.026, 0.025, 0.025, 0.035, 0.035, 0.079, 0.079, 0.072, 0.072, 0.062, 162 | 0.062, 0.107, 0.107, 0.087, 0.087, 0.089, 0.089 163 | ]) 164 | -------------------------------------------------------------------------------- /src/nodes/firecracker_effect_node.py: -------------------------------------------------------------------------------- 1 | from itertools import groupby 2 | from typing import List, Optional, Union 3 | 4 | import cv2 5 | import numpy as np 6 | from mmpose.apis.webcam.nodes import NODES, BaseVisualizerNode 7 | from mmpose.apis.webcam.utils import get_wrist_keypoint_ids 8 | 9 | 10 | @NODES.register_module() 11 | class FirecrackerEffectNode(BaseVisualizerNode): 12 | def __init__(self, 13 | name: str, 14 | input_buffer: str, 15 | output_buffer: Union[str, List[str]], 16 | enable_key: Optional[Union[str, int]] = None, 17 | src_img_path: Optional[str] = None): 18 | 19 | super().__init__(name, input_buffer, output_buffer, enable_key) 20 | 21 | if src_img_path is None: 22 | self.src_img_path = 'https://user-images.githubusercontent' \ 23 | '.com/28900607/149766281-6376055c-ed8b' \ 24 | '-472b-991f-60e6ae6ee1da.gif' 25 | src_img = cv2.VideoCapture(self.src_img_path) 26 | 27 | self.frame_list = [] 28 | ret, frame = src_img.read() 29 | while ret: 30 | self.frame_list.append(frame) 31 | ret, frame = src_img.read() 32 | self.num_frames = len(self.frame_list) 33 | self.frame_idx = 0 34 | self.frame_period = 4 # each frame in gif lasts for 4 frames in video 35 | 36 | @staticmethod 37 | def apply_firecracker_effect(img, 38 | objects, 39 | firecracker_img, 40 | left_wrist_idx, 41 | right_wrist_idx, 42 | kpt_thr=0.5): 43 | """Apply firecracker effect. 44 | Args: 45 | img (np.ndarray): Image data. 46 | objects (list[dict]): The objects with the following information: 47 | - "keypoints" (np.ndarray[K,3]): keypoint detection result in 48 | [x, y, score] 49 | firecracker_img (np.ndarray): Firecracker image with white 50 | background. 51 | left_wrist_idx (int): Keypoint index of left wrist 52 | right_wrist_idx (int): Keypoint index of right wrist 53 | kpt_thr (float): The score threshold of required keypoints. 54 | """ 55 | 56 | hm, wm = firecracker_img.shape[:2] 57 | # anchor points in the firecracker mask 58 | pts_src = np.array([[0. * wm, 0. * hm], [0. * wm, 1. * hm], 59 | [1. * wm, 0. * hm], [1. * wm, 1. * hm]], 60 | dtype=np.float32) 61 | 62 | h, w = img.shape[:2] 63 | h_tar = h / 3 64 | w_tar = h_tar / hm * wm 65 | 66 | for obj in objects: 67 | kpts = obj['keypoints'] 68 | 69 | if kpts[left_wrist_idx, 2] > kpt_thr: 70 | kpt_lwrist = kpts[left_wrist_idx, :2] 71 | # anchor points in the image by eye positions 72 | pts_tar = np.vstack([ 73 | kpt_lwrist - [w_tar / 2, 0], 74 | kpt_lwrist - [w_tar / 2, -h_tar], 75 | kpt_lwrist + [w_tar / 2, 0], 76 | kpt_lwrist + [w_tar / 2, h_tar] 77 | ]) 78 | 79 | h_mat, _ = cv2.findHomography(pts_src, pts_tar) 80 | patch = cv2.warpPerspective(firecracker_img, 81 | h_mat, 82 | dsize=(img.shape[1], img.shape[0]), 83 | borderValue=(255, 255, 255)) 84 | # mask the white background area in the patch with 85 | # a threshold 200 86 | mask = cv2.cvtColor(patch, cv2.COLOR_BGR2GRAY) 87 | mask = (mask < 240).astype(np.uint8) 88 | img = cv2.copyTo(patch, mask, img) 89 | 90 | if kpts[right_wrist_idx, 2] > kpt_thr: 91 | kpt_rwrist = kpts[right_wrist_idx, :2] 92 | 93 | # anchor points in the image by eye positions 94 | pts_tar = np.vstack([ 95 | kpt_rwrist - [w_tar / 2, 0], 96 | kpt_rwrist - [w_tar / 2, -h_tar], 97 | kpt_rwrist + [w_tar / 2, 0], 98 | kpt_rwrist + [w_tar / 2, h_tar] 99 | ]) 100 | 101 | h_mat, _ = cv2.findHomography(pts_src, pts_tar) 102 | patch = cv2.warpPerspective(firecracker_img, 103 | h_mat, 104 | dsize=(img.shape[1], img.shape[0]), 105 | borderValue=(255, 255, 255)) 106 | # mask the white background area in the patch with 107 | # a threshold 200 108 | mask = cv2.cvtColor(patch, cv2.COLOR_BGR2GRAY) 109 | mask = (mask < 240).astype(np.uint8) 110 | img = cv2.copyTo(patch, mask, img) 111 | 112 | return img 113 | 114 | def draw(self, input_msg): 115 | canvas = input_msg.get_image() 116 | 117 | objects = input_msg.get_objects(lambda x: 'keypoints' in x) 118 | frame = self.frame_list[self.frame_idx // self.frame_period] 119 | 120 | for model_cfg, object_group in groupby(objects, 121 | lambda x: x['pose_model_cfg']): 122 | 123 | left_wrist_idx, right_wrist_idx = get_wrist_keypoint_ids(model_cfg) 124 | 125 | canvas = self.apply_firecracker_effect(canvas, object_group, frame, 126 | left_wrist_idx, 127 | right_wrist_idx) 128 | 129 | self.frame_idx = (self.frame_idx + 1) % (self.num_frames * 130 | self.frame_period) 131 | 132 | return canvas 133 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/animalpose.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='animalpose', 3 | paper_info=dict( 4 | author='Cao, Jinkun and Tang, Hongyang and Fang, Hao-Shu and ' 5 | 'Shen, Xiaoyong and Lu, Cewu and Tai, Yu-Wing', 6 | title='Cross-Domain Adaptation for Animal Pose Estimation', 7 | container='The IEEE International Conference on ' 8 | 'Computer Vision (ICCV)', 9 | year='2019', 10 | homepage='https://sites.google.com/view/animal-pose/', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='L_Eye', id=0, color=[0, 255, 0], type='upper', 15 | swap='R_Eye'), 16 | 1: 17 | dict(name='R_Eye', 18 | id=1, 19 | color=[255, 128, 0], 20 | type='upper', 21 | swap='L_Eye'), 22 | 2: 23 | dict(name='L_EarBase', 24 | id=2, 25 | color=[0, 255, 0], 26 | type='upper', 27 | swap='R_EarBase'), 28 | 3: 29 | dict(name='R_EarBase', 30 | id=3, 31 | color=[255, 128, 0], 32 | type='upper', 33 | swap='L_EarBase'), 34 | 4: 35 | dict(name='Nose', id=4, color=[51, 153, 255], type='upper', swap=''), 36 | 5: 37 | dict(name='Throat', id=5, color=[51, 153, 255], type='upper', swap=''), 38 | 6: 39 | dict(name='TailBase', 40 | id=6, 41 | color=[51, 153, 255], 42 | type='lower', 43 | swap=''), 44 | 7: 45 | dict(name='Withers', id=7, color=[51, 153, 255], type='upper', 46 | swap=''), 47 | 8: 48 | dict(name='L_F_Elbow', 49 | id=8, 50 | color=[0, 255, 0], 51 | type='upper', 52 | swap='R_F_Elbow'), 53 | 9: 54 | dict(name='R_F_Elbow', 55 | id=9, 56 | color=[255, 128, 0], 57 | type='upper', 58 | swap='L_F_Elbow'), 59 | 10: 60 | dict(name='L_B_Elbow', 61 | id=10, 62 | color=[0, 255, 0], 63 | type='lower', 64 | swap='R_B_Elbow'), 65 | 11: 66 | dict(name='R_B_Elbow', 67 | id=11, 68 | color=[255, 128, 0], 69 | type='lower', 70 | swap='L_B_Elbow'), 71 | 12: 72 | dict(name='L_F_Knee', 73 | id=12, 74 | color=[0, 255, 0], 75 | type='upper', 76 | swap='R_F_Knee'), 77 | 13: 78 | dict(name='R_F_Knee', 79 | id=13, 80 | color=[255, 128, 0], 81 | type='upper', 82 | swap='L_F_Knee'), 83 | 14: 84 | dict(name='L_B_Knee', 85 | id=14, 86 | color=[0, 255, 0], 87 | type='lower', 88 | swap='R_B_Knee'), 89 | 15: 90 | dict(name='R_B_Knee', 91 | id=15, 92 | color=[255, 128, 0], 93 | type='lower', 94 | swap='L_B_Knee'), 95 | 16: 96 | dict(name='L_F_Paw', 97 | id=16, 98 | color=[0, 255, 0], 99 | type='upper', 100 | swap='R_F_Paw'), 101 | 17: 102 | dict(name='R_F_Paw', 103 | id=17, 104 | color=[255, 128, 0], 105 | type='upper', 106 | swap='L_F_Paw'), 107 | 18: 108 | dict(name='L_B_Paw', 109 | id=18, 110 | color=[0, 255, 0], 111 | type='lower', 112 | swap='R_B_Paw'), 113 | 19: 114 | dict(name='R_B_Paw', 115 | id=19, 116 | color=[255, 128, 0], 117 | type='lower', 118 | swap='L_B_Paw') 119 | }, 120 | skeleton_info={ 121 | 0: dict(link=('L_Eye', 'R_Eye'), id=0, color=[51, 153, 255]), 122 | 1: dict(link=('L_Eye', 'L_EarBase'), id=1, color=[0, 255, 0]), 123 | 2: dict(link=('R_Eye', 'R_EarBase'), id=2, color=[255, 128, 0]), 124 | 3: dict(link=('L_Eye', 'Nose'), id=3, color=[0, 255, 0]), 125 | 4: dict(link=('R_Eye', 'Nose'), id=4, color=[255, 128, 0]), 126 | 5: dict(link=('Nose', 'Throat'), id=5, color=[51, 153, 255]), 127 | 6: dict(link=('Throat', 'Withers'), id=6, color=[51, 153, 255]), 128 | 7: dict(link=('TailBase', 'Withers'), id=7, color=[51, 153, 255]), 129 | 8: dict(link=('Throat', 'L_F_Elbow'), id=8, color=[0, 255, 0]), 130 | 9: dict(link=('L_F_Elbow', 'L_F_Knee'), id=9, color=[0, 255, 0]), 131 | 10: dict(link=('L_F_Knee', 'L_F_Paw'), id=10, color=[0, 255, 0]), 132 | 11: dict(link=('Throat', 'R_F_Elbow'), id=11, color=[255, 128, 0]), 133 | 12: dict(link=('R_F_Elbow', 'R_F_Knee'), id=12, color=[255, 128, 0]), 134 | 13: dict(link=('R_F_Knee', 'R_F_Paw'), id=13, color=[255, 128, 0]), 135 | 14: dict(link=('TailBase', 'L_B_Elbow'), id=14, color=[0, 255, 0]), 136 | 15: dict(link=('L_B_Elbow', 'L_B_Knee'), id=15, color=[0, 255, 0]), 137 | 16: dict(link=('L_B_Knee', 'L_B_Paw'), id=16, color=[0, 255, 0]), 138 | 17: dict(link=('TailBase', 'R_B_Elbow'), id=17, color=[255, 128, 0]), 139 | 18: dict(link=('R_B_Elbow', 'R_B_Knee'), id=18, color=[255, 128, 0]), 140 | 19: dict(link=('R_B_Knee', 'R_B_Paw'), id=19, color=[255, 128, 0]) 141 | }, 142 | joint_weights=[ 143 | 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.2, 1.2, 144 | 1.5, 1.5, 1.5, 1.5 145 | ], 146 | 147 | # Note: The original paper did not provide enough information about 148 | # the sigmas. We modified from 'https://github.com/cocodataset/' 149 | # 'cocoapi/blob/master/PythonAPI/pycocotools/cocoeval.py#L523' 150 | sigmas=[ 151 | 0.025, 0.025, 0.026, 0.035, 0.035, 0.10, 0.10, 0.10, 0.107, 0.107, 152 | 0.107, 0.107, 0.087, 0.087, 0.087, 0.087, 0.089, 0.089, 0.089, 0.089 153 | ]) 154 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/macaque.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='macaque', 3 | paper_info=dict( 4 | author='Labuguen, Rollyn and Matsumoto, Jumpei and ' 5 | 'Negrete, Salvador and Nishimaru, Hiroshi and ' 6 | 'Nishijo, Hisao and Takada, Masahiko and ' 7 | 'Go, Yasuhiro and Inoue, Ken-ichi and Shibata, Tomohiro', 8 | title='MacaquePose: A novel "in the wild" macaque monkey pose dataset ' 9 | 'for markerless motion capture', 10 | container='bioRxiv', 11 | year='2020', 12 | homepage='http://www.pri.kyoto-u.ac.jp/datasets/' 13 | 'macaquepose/index.html', 14 | ), 15 | keypoint_info={ 16 | 0: 17 | dict(name='nose', id=0, color=[51, 153, 255], type='upper', swap=''), 18 | 1: 19 | dict(name='left_eye', 20 | id=1, 21 | color=[51, 153, 255], 22 | type='upper', 23 | swap='right_eye'), 24 | 2: 25 | dict(name='right_eye', 26 | id=2, 27 | color=[51, 153, 255], 28 | type='upper', 29 | swap='left_eye'), 30 | 3: 31 | dict(name='left_ear', 32 | id=3, 33 | color=[51, 153, 255], 34 | type='upper', 35 | swap='right_ear'), 36 | 4: 37 | dict(name='right_ear', 38 | id=4, 39 | color=[51, 153, 255], 40 | type='upper', 41 | swap='left_ear'), 42 | 5: 43 | dict(name='left_shoulder', 44 | id=5, 45 | color=[0, 255, 0], 46 | type='upper', 47 | swap='right_shoulder'), 48 | 6: 49 | dict(name='right_shoulder', 50 | id=6, 51 | color=[255, 128, 0], 52 | type='upper', 53 | swap='left_shoulder'), 54 | 7: 55 | dict(name='left_elbow', 56 | id=7, 57 | color=[0, 255, 0], 58 | type='upper', 59 | swap='right_elbow'), 60 | 8: 61 | dict(name='right_elbow', 62 | id=8, 63 | color=[255, 128, 0], 64 | type='upper', 65 | swap='left_elbow'), 66 | 9: 67 | dict(name='left_wrist', 68 | id=9, 69 | color=[0, 255, 0], 70 | type='upper', 71 | swap='right_wrist'), 72 | 10: 73 | dict(name='right_wrist', 74 | id=10, 75 | color=[255, 128, 0], 76 | type='upper', 77 | swap='left_wrist'), 78 | 11: 79 | dict(name='left_hip', 80 | id=11, 81 | color=[0, 255, 0], 82 | type='lower', 83 | swap='right_hip'), 84 | 12: 85 | dict(name='right_hip', 86 | id=12, 87 | color=[255, 128, 0], 88 | type='lower', 89 | swap='left_hip'), 90 | 13: 91 | dict(name='left_knee', 92 | id=13, 93 | color=[0, 255, 0], 94 | type='lower', 95 | swap='right_knee'), 96 | 14: 97 | dict(name='right_knee', 98 | id=14, 99 | color=[255, 128, 0], 100 | type='lower', 101 | swap='left_knee'), 102 | 15: 103 | dict(name='left_ankle', 104 | id=15, 105 | color=[0, 255, 0], 106 | type='lower', 107 | swap='right_ankle'), 108 | 16: 109 | dict(name='right_ankle', 110 | id=16, 111 | color=[255, 128, 0], 112 | type='lower', 113 | swap='left_ankle') 114 | }, 115 | skeleton_info={ 116 | 0: 117 | dict(link=('left_ankle', 'left_knee'), id=0, color=[0, 255, 0]), 118 | 1: 119 | dict(link=('left_knee', 'left_hip'), id=1, color=[0, 255, 0]), 120 | 2: 121 | dict(link=('right_ankle', 'right_knee'), id=2, color=[255, 128, 0]), 122 | 3: 123 | dict(link=('right_knee', 'right_hip'), id=3, color=[255, 128, 0]), 124 | 4: 125 | dict(link=('left_hip', 'right_hip'), id=4, color=[51, 153, 255]), 126 | 5: 127 | dict(link=('left_shoulder', 'left_hip'), id=5, color=[51, 153, 255]), 128 | 6: 129 | dict(link=('right_shoulder', 'right_hip'), id=6, color=[51, 153, 255]), 130 | 7: 131 | dict(link=('left_shoulder', 'right_shoulder'), 132 | id=7, 133 | color=[51, 153, 255]), 134 | 8: 135 | dict(link=('left_shoulder', 'left_elbow'), id=8, color=[0, 255, 0]), 136 | 9: 137 | dict(link=('right_shoulder', 'right_elbow'), id=9, color=[255, 128, 138 | 0]), 139 | 10: 140 | dict(link=('left_elbow', 'left_wrist'), id=10, color=[0, 255, 0]), 141 | 11: 142 | dict(link=('right_elbow', 'right_wrist'), id=11, color=[255, 128, 0]), 143 | 12: 144 | dict(link=('left_eye', 'right_eye'), id=12, color=[51, 153, 255]), 145 | 13: 146 | dict(link=('nose', 'left_eye'), id=13, color=[51, 153, 255]), 147 | 14: 148 | dict(link=('nose', 'right_eye'), id=14, color=[51, 153, 255]), 149 | 15: 150 | dict(link=('left_eye', 'left_ear'), id=15, color=[51, 153, 255]), 151 | 16: 152 | dict(link=('right_eye', 'right_ear'), id=16, color=[51, 153, 255]), 153 | 17: 154 | dict(link=('left_ear', 'left_shoulder'), id=17, color=[51, 153, 255]), 155 | 18: 156 | dict(link=('right_ear', 'right_shoulder'), id=18, color=[51, 153, 255]) 157 | }, 158 | joint_weights=[ 159 | 1., 1., 1., 1., 1., 1., 1., 1.2, 1.2, 1.5, 1.5, 1., 1., 1.2, 1.2, 1.5, 160 | 1.5 161 | ], 162 | sigmas=[ 163 | 0.026, 0.025, 0.025, 0.035, 0.035, 0.079, 0.079, 0.072, 0.072, 0.062, 164 | 0.062, 0.107, 0.107, 0.087, 0.087, 0.089, 0.089 165 | ]) 166 | -------------------------------------------------------------------------------- /configs/examples/pose_estimation/pose_estimation.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) OpenMMLab. All rights reserved. 2 | executor_cfg = dict( 3 | # Basic configurations of the executor 4 | name='Pose Estimation', 5 | camera_id=0, 6 | # Define nodes. 7 | # The configuration of a node usually includes: 8 | # 1. 'type': Node class name 9 | # 2. 'name': Node name 10 | # 3. I/O buffers (e.g. 'input_buffer', 'output_buffer'): specify the 11 | # input and output buffer names. This may depend on the node class. 12 | # 4. 'enable_key': assign a hot-key to toggle enable/disable this node. 13 | # This may depend on the node class. 14 | # 5. Other class-specific arguments 15 | nodes=[ 16 | # 'DetectorNode': 17 | # This node performs object detection from the frame image using an 18 | # MMDetection model. 19 | dict( 20 | type='DetectorNode', 21 | name='detector', 22 | model_config='model_configs/mmdet/' 23 | 'ssdlite_mobilenetv2_scratch_600e_coco.py', 24 | model_checkpoint='https://download.openmmlab.com' 25 | '/mmdetection/v2.0/ssd/' 26 | 'ssdlite_mobilenetv2_scratch_600e_coco/ssdlite_mobilenetv2_' 27 | 'scratch_600e_coco_20210629_110627-974d9307.pth', 28 | input_buffer='_input_', # `_input_` is an executor-reserved buffer 29 | output_buffer='det_result'), 30 | # 'TopDownPoseEstimatorNode': 31 | # This node performs keypoint detection from the frame image using an 32 | # MMPose top-down model. Detection results is needed. 33 | dict(type='TopDownPoseEstimatorNode', 34 | name='human pose estimator', 35 | model_config='model_configs/mmpose/' 36 | 'vipnas_mbv3_coco_wholebody_256x192_dark.py', 37 | model_checkpoint='https://download.openmmlab.com/mmpose/top_down/' 38 | 'vipnas/vipnas_mbv3_coco_wholebody_256x192_dark' 39 | '-e2158108_20211205.pth', 40 | labels=['person'], 41 | smooth=True, 42 | input_buffer='det_result', 43 | output_buffer='human_pose'), 44 | dict(type='TopDownPoseEstimatorNode', 45 | name='animal pose estimator', 46 | model_config='model_configs/' 47 | 'mmpose/hrnet_w32_animalpose_256x256.py', 48 | model_checkpoint='https://download.openmmlab.com/mmpose/animal/' 49 | 'hrnet/hrnet_w32_animalpose_256x256-1aa7f075_20210426.pth', 50 | labels=['cat', 'dog', 'horse', 'sheep', 'cow'], 51 | input_buffer='human_pose', 52 | output_buffer='animal_pose'), 53 | # 'ObjectAssignerNode': 54 | # This node binds the latest model inference result with the current 55 | # frame. (This means the frame image and inference result may be 56 | # asynchronous). 57 | dict( 58 | type='ObjectAssignerNode', 59 | name='object assigner', 60 | frame_buffer='_frame_', # `_frame_` is an executor-reserved buffer 61 | object_buffer='animal_pose', 62 | output_buffer='frame'), 63 | # 'ObjectVisualizerNode': 64 | # This node draw the pose visualization result in the frame image. 65 | # Pose results is needed. 66 | dict(type='ObjectVisualizerNode', 67 | name='object visualizer', 68 | enable_key='v', 69 | enable=True, 70 | show_bbox=True, 71 | must_have_keypoint=False, 72 | show_keypoint=True, 73 | input_buffer='frame', 74 | output_buffer='vis'), 75 | # 'SunglassesNode': 76 | # This node draw the sunglasses effect in the frame image. 77 | # Pose results is needed. 78 | dict(type='SunglassesEffectNode', 79 | name='sunglasses', 80 | enable_key='s', 81 | enable=False, 82 | input_buffer='vis', 83 | output_buffer='vis_sunglasses'), 84 | # 'BigeyeEffectNode': 85 | # This node draw the big-eye effetc in the frame image. 86 | # Pose results is needed. 87 | dict(type='BigeyeEffectNode', 88 | name='big-eye', 89 | enable_key='b', 90 | enable=False, 91 | input_buffer='vis_sunglasses', 92 | output_buffer='vis_bigeye'), 93 | # 'NoticeBoardNode': 94 | # This node show a notice board with given content, e.g. help 95 | # information. 96 | dict( 97 | type='NoticeBoardNode', 98 | name='instruction', 99 | enable_key='h', 100 | enable=True, 101 | input_buffer='vis_bigeye', 102 | output_buffer='vis_notice', 103 | content_lines=[ 104 | 'This is a demo for pose visualization and simple image ' 105 | 'effects. Have fun!', '', 'Hot-keys:', 106 | '"v": Pose estimation result visualization', 107 | '"s": Sunglasses effect B-)', '"b": Bug-eye effect 0_0', 108 | '"h": Show help information', 109 | '"m": Show diagnostic information', '"q": Exit' 110 | ], 111 | ), 112 | # 'MonitorNode': 113 | # This node show diagnostic information in the frame image. It can 114 | # be used for debugging or monitoring system resource status. 115 | dict(type='MonitorNode', 116 | name='monitor', 117 | enable_key='m', 118 | enable=False, 119 | input_buffer='vis_notice', 120 | output_buffer='display'), 121 | # 'RecorderNode': 122 | # This node save the output video into a file. 123 | dict(type='RecorderNode', 124 | name='recorder', 125 | out_video_file='webcam_demo.mp4', 126 | input_buffer='display', 127 | output_buffer='_display_' 128 | # `_display_` is an executor-reserved buffer 129 | ) 130 | ]) 131 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/rhd2d.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='rhd2d', 3 | paper_info=dict( 4 | author='Christian Zimmermann and Thomas Brox', 5 | title='Learning to Estimate 3D Hand Pose from Single RGB Images', 6 | container='arXiv', 7 | year='2017', 8 | homepage='https://lmb.informatik.uni-freiburg.de/resources/' 9 | 'datasets/RenderedHandposeDataset.en.html', 10 | ), 11 | # In RHD, 1-4: left thumb [tip to palm], which means the finger is from 12 | # tip to palm, so as other fingers. Please refer to 13 | # `https://lmb.informatik.uni-freiburg.de/resources/datasets/ 14 | # RenderedHandpose/README` for details of keypoint definition. 15 | # But in COCO-WholeBody-Hand, FreiHand, CMU Panoptic HandDB, it is in 16 | # inverse order. Pay attention to this if you want to combine RHD with 17 | # other hand datasets to train a single model. 18 | # Also, note that 'keypoint_info' will not directly affect the order of 19 | # the keypoint in the dataset. It is mostly for visualization & storing 20 | # information about flip_pairs. 21 | keypoint_info={ 22 | 0: 23 | dict(name='wrist', id=0, color=[255, 255, 255], type='', swap=''), 24 | 1: 25 | dict(name='thumb4', id=1, color=[255, 128, 0], type='', swap=''), 26 | 2: 27 | dict(name='thumb3', id=2, color=[255, 128, 0], type='', swap=''), 28 | 3: 29 | dict(name='thumb2', id=3, color=[255, 128, 0], type='', swap=''), 30 | 4: 31 | dict(name='thumb1', id=4, color=[255, 128, 0], type='', swap=''), 32 | 5: 33 | dict(name='forefinger4', id=5, color=[255, 153, 255], type='', 34 | swap=''), 35 | 6: 36 | dict(name='forefinger3', id=6, color=[255, 153, 255], type='', 37 | swap=''), 38 | 7: 39 | dict(name='forefinger2', id=7, color=[255, 153, 255], type='', 40 | swap=''), 41 | 8: 42 | dict(name='forefinger1', id=8, color=[255, 153, 255], type='', 43 | swap=''), 44 | 9: 45 | dict(name='middle_finger4', 46 | id=9, 47 | color=[102, 178, 255], 48 | type='', 49 | swap=''), 50 | 10: 51 | dict(name='middle_finger3', 52 | id=10, 53 | color=[102, 178, 255], 54 | type='', 55 | swap=''), 56 | 11: 57 | dict(name='middle_finger2', 58 | id=11, 59 | color=[102, 178, 255], 60 | type='', 61 | swap=''), 62 | 12: 63 | dict(name='middle_finger1', 64 | id=12, 65 | color=[102, 178, 255], 66 | type='', 67 | swap=''), 68 | 13: 69 | dict(name='ring_finger4', id=13, color=[255, 51, 51], type='', 70 | swap=''), 71 | 14: 72 | dict(name='ring_finger3', id=14, color=[255, 51, 51], type='', 73 | swap=''), 74 | 15: 75 | dict(name='ring_finger2', id=15, color=[255, 51, 51], type='', 76 | swap=''), 77 | 16: 78 | dict(name='ring_finger1', id=16, color=[255, 51, 51], type='', 79 | swap=''), 80 | 17: 81 | dict(name='pinky_finger4', id=17, color=[0, 255, 0], type='', swap=''), 82 | 18: 83 | dict(name='pinky_finger3', id=18, color=[0, 255, 0], type='', swap=''), 84 | 19: 85 | dict(name='pinky_finger2', id=19, color=[0, 255, 0], type='', swap=''), 86 | 20: 87 | dict(name='pinky_finger1', id=20, color=[0, 255, 0], type='', swap='') 88 | }, 89 | skeleton_info={ 90 | 0: 91 | dict(link=('wrist', 'thumb1'), id=0, color=[255, 128, 0]), 92 | 1: 93 | dict(link=('thumb1', 'thumb2'), id=1, color=[255, 128, 0]), 94 | 2: 95 | dict(link=('thumb2', 'thumb3'), id=2, color=[255, 128, 0]), 96 | 3: 97 | dict(link=('thumb3', 'thumb4'), id=3, color=[255, 128, 0]), 98 | 4: 99 | dict(link=('wrist', 'forefinger1'), id=4, color=[255, 153, 255]), 100 | 5: 101 | dict(link=('forefinger1', 'forefinger2'), id=5, color=[255, 153, 255]), 102 | 6: 103 | dict(link=('forefinger2', 'forefinger3'), id=6, color=[255, 153, 255]), 104 | 7: 105 | dict(link=('forefinger3', 'forefinger4'), id=7, color=[255, 153, 255]), 106 | 8: 107 | dict(link=('wrist', 'middle_finger1'), id=8, color=[102, 178, 255]), 108 | 9: 109 | dict(link=('middle_finger1', 'middle_finger2'), 110 | id=9, 111 | color=[102, 178, 255]), 112 | 10: 113 | dict(link=('middle_finger2', 'middle_finger3'), 114 | id=10, 115 | color=[102, 178, 255]), 116 | 11: 117 | dict(link=('middle_finger3', 'middle_finger4'), 118 | id=11, 119 | color=[102, 178, 255]), 120 | 12: 121 | dict(link=('wrist', 'ring_finger1'), id=12, color=[255, 51, 51]), 122 | 13: 123 | dict(link=('ring_finger1', 'ring_finger2'), id=13, color=[255, 51, 124 | 51]), 125 | 14: 126 | dict(link=('ring_finger2', 'ring_finger3'), id=14, color=[255, 51, 127 | 51]), 128 | 15: 129 | dict(link=('ring_finger3', 'ring_finger4'), id=15, color=[255, 51, 130 | 51]), 131 | 16: 132 | dict(link=('wrist', 'pinky_finger1'), id=16, color=[0, 255, 0]), 133 | 17: 134 | dict(link=('pinky_finger1', 'pinky_finger2'), id=17, color=[0, 255, 135 | 0]), 136 | 18: 137 | dict(link=('pinky_finger2', 'pinky_finger3'), id=18, color=[0, 255, 138 | 0]), 139 | 19: 140 | dict(link=('pinky_finger3', 'pinky_finger4'), id=19, color=[0, 255, 0]) 141 | }, 142 | joint_weights=[1.] * 21, 143 | sigmas=[]) 144 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/horse10.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='horse10', 3 | paper_info=dict( 4 | author='Mathis, Alexander and Biasi, Thomas and ' 5 | 'Schneider, Steffen and ' 6 | 'Yuksekgonul, Mert and Rogers, Byron and ' 7 | 'Bethge, Matthias and ' 8 | 'Mathis, Mackenzie W', 9 | title='Pretraining boosts out-of-domain robustness ' 10 | 'for pose estimation', 11 | container='Proceedings of the IEEE/CVF Winter Conference on ' 12 | 'Applications of Computer Vision', 13 | year='2021', 14 | homepage='http://www.mackenziemathislab.org/horse10', 15 | ), 16 | keypoint_info={ 17 | 0: 18 | dict(name='Nose', id=0, color=[255, 153, 255], type='upper', swap=''), 19 | 1: 20 | dict(name='Eye', id=1, color=[255, 153, 255], type='upper', swap=''), 21 | 2: 22 | dict(name='Nearknee', 23 | id=2, 24 | color=[255, 102, 255], 25 | type='upper', 26 | swap=''), 27 | 3: 28 | dict(name='Nearfrontfetlock', 29 | id=3, 30 | color=[255, 102, 255], 31 | type='upper', 32 | swap=''), 33 | 4: 34 | dict(name='Nearfrontfoot', 35 | id=4, 36 | color=[255, 102, 255], 37 | type='upper', 38 | swap=''), 39 | 5: 40 | dict(name='Offknee', 41 | id=5, 42 | color=[255, 102, 255], 43 | type='upper', 44 | swap=''), 45 | 6: 46 | dict(name='Offfrontfetlock', 47 | id=6, 48 | color=[255, 102, 255], 49 | type='upper', 50 | swap=''), 51 | 7: 52 | dict(name='Offfrontfoot', 53 | id=7, 54 | color=[255, 102, 255], 55 | type='upper', 56 | swap=''), 57 | 8: 58 | dict(name='Shoulder', 59 | id=8, 60 | color=[255, 153, 255], 61 | type='upper', 62 | swap=''), 63 | 9: 64 | dict(name='Midshoulder', 65 | id=9, 66 | color=[255, 153, 255], 67 | type='upper', 68 | swap=''), 69 | 10: 70 | dict(name='Elbow', id=10, color=[255, 153, 255], type='upper', 71 | swap=''), 72 | 11: 73 | dict(name='Girth', id=11, color=[255, 153, 255], type='upper', 74 | swap=''), 75 | 12: 76 | dict(name='Wither', 77 | id=12, 78 | color=[255, 153, 255], 79 | type='upper', 80 | swap=''), 81 | 13: 82 | dict(name='Nearhindhock', 83 | id=13, 84 | color=[255, 51, 255], 85 | type='lower', 86 | swap=''), 87 | 14: 88 | dict(name='Nearhindfetlock', 89 | id=14, 90 | color=[255, 51, 255], 91 | type='lower', 92 | swap=''), 93 | 15: 94 | dict(name='Nearhindfoot', 95 | id=15, 96 | color=[255, 51, 255], 97 | type='lower', 98 | swap=''), 99 | 16: 100 | dict(name='Hip', id=16, color=[255, 153, 255], type='lower', swap=''), 101 | 17: 102 | dict(name='Stifle', 103 | id=17, 104 | color=[255, 153, 255], 105 | type='lower', 106 | swap=''), 107 | 18: 108 | dict(name='Offhindhock', 109 | id=18, 110 | color=[255, 51, 255], 111 | type='lower', 112 | swap=''), 113 | 19: 114 | dict(name='Offhindfetlock', 115 | id=19, 116 | color=[255, 51, 255], 117 | type='lower', 118 | swap=''), 119 | 20: 120 | dict(name='Offhindfoot', 121 | id=20, 122 | color=[255, 51, 255], 123 | type='lower', 124 | swap=''), 125 | 21: 126 | dict(name='Ischium', 127 | id=21, 128 | color=[255, 153, 255], 129 | type='lower', 130 | swap='') 131 | }, 132 | skeleton_info={ 133 | 0: 134 | dict(link=('Nose', 'Eye'), id=0, color=[255, 153, 255]), 135 | 1: 136 | dict(link=('Eye', 'Wither'), id=1, color=[255, 153, 255]), 137 | 2: 138 | dict(link=('Wither', 'Hip'), id=2, color=[255, 153, 255]), 139 | 3: 140 | dict(link=('Hip', 'Ischium'), id=3, color=[255, 153, 255]), 141 | 4: 142 | dict(link=('Ischium', 'Stifle'), id=4, color=[255, 153, 255]), 143 | 5: 144 | dict(link=('Stifle', 'Girth'), id=5, color=[255, 153, 255]), 145 | 6: 146 | dict(link=('Girth', 'Elbow'), id=6, color=[255, 153, 255]), 147 | 7: 148 | dict(link=('Elbow', 'Shoulder'), id=7, color=[255, 153, 255]), 149 | 8: 150 | dict(link=('Shoulder', 'Midshoulder'), id=8, color=[255, 153, 255]), 151 | 9: 152 | dict(link=('Midshoulder', 'Wither'), id=9, color=[255, 153, 255]), 153 | 10: 154 | dict(link=('Nearknee', 'Nearfrontfetlock'), 155 | id=10, 156 | color=[255, 102, 255]), 157 | 11: 158 | dict(link=('Nearfrontfetlock', 'Nearfrontfoot'), 159 | id=11, 160 | color=[255, 102, 255]), 161 | 12: 162 | dict(link=('Offknee', 'Offfrontfetlock'), id=12, color=[255, 102, 163 | 255]), 164 | 13: 165 | dict(link=('Offfrontfetlock', 'Offfrontfoot'), 166 | id=13, 167 | color=[255, 102, 255]), 168 | 14: 169 | dict(link=('Nearhindhock', 'Nearhindfetlock'), 170 | id=14, 171 | color=[255, 51, 255]), 172 | 15: 173 | dict(link=('Nearhindfetlock', 'Nearhindfoot'), 174 | id=15, 175 | color=[255, 51, 255]), 176 | 16: 177 | dict(link=('Offhindhock', 'Offhindfetlock'), 178 | id=16, 179 | color=[255, 51, 255]), 180 | 17: 181 | dict(link=('Offhindfetlock', 'Offhindfoot'), 182 | id=17, 183 | color=[255, 51, 255]) 184 | }, 185 | joint_weights=[1.] * 22, 186 | sigmas=[]) 187 | -------------------------------------------------------------------------------- /model_configs/mmpose/hrnet_w32_animalpose_256x256.py: -------------------------------------------------------------------------------- 1 | _base_ = ['./_base_/default_runtime.py', './_base_/datasets/animalpose.py'] 2 | evaluation = dict(interval=10, metric='mAP', save_best='AP') 3 | 4 | optimizer = dict( 5 | type='Adam', 6 | lr=5e-4, 7 | ) 8 | optimizer_config = dict(grad_clip=None) 9 | # learning policy 10 | lr_config = dict(policy='step', 11 | warmup='linear', 12 | warmup_iters=500, 13 | warmup_ratio=0.001, 14 | step=[170, 200]) 15 | total_epochs = 210 16 | log_config = dict( 17 | interval=1, 18 | hooks=[ 19 | dict(type='TextLoggerHook'), 20 | # dict(type='TensorboardLoggerHook') 21 | ]) 22 | 23 | channel_cfg = dict( 24 | num_output_channels=20, 25 | dataset_joints=20, 26 | dataset_channel=[ 27 | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19], 28 | ], 29 | inference_channel=[ 30 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 31 | ]) 32 | 33 | # model settings 34 | model = dict(type='TopDown', 35 | pretrained='https://download.openmmlab.com/mmpose/' 36 | 'pretrain_models/hrnet_w32-36af842e.pth', 37 | backbone=dict( 38 | type='HRNet', 39 | in_channels=3, 40 | extra=dict(stage1=dict(num_modules=1, 41 | num_branches=1, 42 | block='BOTTLENECK', 43 | num_blocks=(4, ), 44 | num_channels=(64, )), 45 | stage2=dict(num_modules=1, 46 | num_branches=2, 47 | block='BASIC', 48 | num_blocks=(4, 4), 49 | num_channels=(32, 64)), 50 | stage3=dict(num_modules=4, 51 | num_branches=3, 52 | block='BASIC', 53 | num_blocks=(4, 4, 4), 54 | num_channels=(32, 64, 128)), 55 | stage4=dict(num_modules=3, 56 | num_branches=4, 57 | block='BASIC', 58 | num_blocks=(4, 4, 4, 4), 59 | num_channels=(32, 64, 128, 256))), 60 | ), 61 | keypoint_head=dict( 62 | type='TopdownHeatmapSimpleHead', 63 | in_channels=32, 64 | out_channels=channel_cfg['num_output_channels'], 65 | num_deconv_layers=0, 66 | extra=dict(final_conv_kernel=1, ), 67 | loss_keypoint=dict(type='JointsMSELoss', 68 | use_target_weight=True)), 69 | train_cfg=dict(), 70 | test_cfg=dict(flip_test=True, 71 | post_process='default', 72 | shift_heatmap=True, 73 | modulate_kernel=11)) 74 | 75 | data_cfg = dict( 76 | image_size=[256, 256], 77 | heatmap_size=[64, 64], 78 | num_output_channels=channel_cfg['num_output_channels'], 79 | num_joints=channel_cfg['dataset_joints'], 80 | dataset_channel=channel_cfg['dataset_channel'], 81 | inference_channel=channel_cfg['inference_channel'], 82 | soft_nms=False, 83 | nms_thr=1.0, 84 | oks_thr=0.9, 85 | vis_thr=0.2, 86 | use_gt_bbox=True, 87 | det_bbox_thr=0.0, 88 | bbox_file='', 89 | ) 90 | 91 | train_pipeline = [ 92 | dict(type='LoadImageFromFile'), 93 | dict(type='TopDownGetBboxCenterScale', padding=1.25), 94 | dict(type='TopDownRandomShiftBboxCenter', shift_factor=0.16, prob=0.3), 95 | dict(type='TopDownRandomFlip', flip_prob=0.5), 96 | dict(type='TopDownHalfBodyTransform', 97 | num_joints_half_body=8, 98 | prob_half_body=0.3), 99 | dict(type='TopDownGetRandomScaleRotation', rot_factor=40, 100 | scale_factor=0.5), 101 | dict(type='TopDownAffine'), 102 | dict(type='ToTensor'), 103 | dict(type='NormalizeTensor', 104 | mean=[0.485, 0.456, 0.406], 105 | std=[0.229, 0.224, 0.225]), 106 | dict(type='TopDownGenerateTarget', sigma=2), 107 | dict(type='Collect', 108 | keys=['img', 'target', 'target_weight'], 109 | meta_keys=[ 110 | 'image_file', 'joints_3d', 'joints_3d_visible', 'center', 'scale', 111 | 'rotation', 'bbox_score', 'flip_pairs' 112 | ]), 113 | ] 114 | 115 | val_pipeline = [ 116 | dict(type='LoadImageFromFile'), 117 | dict(type='TopDownGetBboxCenterScale', padding=1.25), 118 | dict(type='TopDownAffine'), 119 | dict(type='ToTensor'), 120 | dict(type='NormalizeTensor', 121 | mean=[0.485, 0.456, 0.406], 122 | std=[0.229, 0.224, 0.225]), 123 | dict(type='Collect', 124 | keys=['img'], 125 | meta_keys=[ 126 | 'image_file', 'center', 'scale', 'rotation', 'bbox_score', 127 | 'flip_pairs' 128 | ]), 129 | ] 130 | 131 | test_pipeline = val_pipeline 132 | 133 | data_root = 'data/animalpose' 134 | data = dict( 135 | samples_per_gpu=64, 136 | workers_per_gpu=2, 137 | val_dataloader=dict(samples_per_gpu=32), 138 | test_dataloader=dict(samples_per_gpu=32), 139 | train=dict(type='AnimalPoseDataset', 140 | ann_file=f'{data_root}/annotations/animalpose_train.json', 141 | img_prefix=f'{data_root}/', 142 | data_cfg=data_cfg, 143 | pipeline=train_pipeline, 144 | dataset_info={{_base_.dataset_info}}), 145 | val=dict(type='AnimalPoseDataset', 146 | ann_file=f'{data_root}/annotations/animalpose_val.json', 147 | img_prefix=f'{data_root}/', 148 | data_cfg=data_cfg, 149 | pipeline=val_pipeline, 150 | dataset_info={{_base_.dataset_info}}), 151 | test=dict(type='AnimalPoseDataset', 152 | ann_file=f'{data_root}/annotations/animalpose_val.json', 153 | img_prefix=f'{data_root}/', 154 | data_cfg=data_cfg, 155 | pipeline=test_pipeline, 156 | dataset_info={{_base_.dataset_info}}), 157 | ) 158 | -------------------------------------------------------------------------------- /configs/examples/bing_dwen_dwen/resource-info.json: -------------------------------------------------------------------------------- 1 | [ 2 | { 3 | "id": 1, 4 | "result": "{\"width\":690,\"height\":713,\"valid\":true,\"rotate\":0,\"step_1\":{\"toolName\":\"pointTool\",\"result\":[{\"x\":374.86387434554973,\"y\":262.8020942408377,\"attribute\":\"\",\"valid\":true,\"id\":\"8SK9cVyu\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":2},{\"x\":492.8261780104712,\"y\":285.2,\"attribute\":\"\",\"valid\":true,\"id\":\"qDk54WsI\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":1},{\"x\":430.11204188481673,\"y\":318.0502617801047,\"attribute\":\"\",\"valid\":true,\"id\":\"4H80L7lL\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":3}]},\"step_2\":{\"dataSourceStep\":0,\"toolName\":\"polygonTool\",\"result\":[{\"id\":\"pwUsrf9u\",\"sourceID\":\"\",\"valid\":true,\"textAttribute\":\"\",\"pointList\":[{\"x\":423.3926701570681,\"y\":191.87539267015708},{\"x\":488.3465968586388,\"y\":209.04712041884818},{\"x\":535.3821989528797,\"y\":248.6167539267016},{\"x\":549.5675392670157,\"y\":306.8513089005236},{\"x\":537.6219895287959,\"y\":349.407329842932},{\"x\":510.74450261780106,\"y\":381.51099476439794},{\"x\":480.1340314136126,\"y\":394.9497382198953},{\"x\":411.4471204188482,\"y\":390.47015706806286},{\"x\":355.45235602094243,\"y\":373.29842931937173},{\"x\":306.17696335078534,\"y\":327.00942408376966},{\"x\":294.97801047120424,\"y\":284.45340314136126},{\"x\":306.9235602094241,\"y\":245.6303664921466},{\"x\":333.8010471204189,\"y\":217.25968586387435},{\"x\":370.3842931937173,\"y\":196.35497382198955}],\"attribute\":\"\",\"order\":1}]}}", 5 | "url": "https://user-images.githubusercontent.com/15977946/152742677-35fe8a01-bd06-4a12-a02e-949e7d71f28a.jpg", 6 | "fileName": "bing_dwen_dwen1.jpg" 7 | }, 8 | { 9 | "id": 2, 10 | "result": "{\"width\":690,\"height\":659,\"valid\":true,\"rotate\":0,\"step_1\":{\"dataSourceStep\":0,\"toolName\":\"pointTool\",\"result\":[{\"x\":293.2460732984293,\"y\":242.89842931937173,\"attribute\":\"\",\"valid\":true,\"id\":\"KgPs39bY\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":1},{\"x\":170.41675392670155,\"y\":270.50052356020944,\"attribute\":\"\",\"valid\":true,\"id\":\"XwHyoBFU\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":2},{\"x\":224.24083769633506,\"y\":308.45340314136126,\"attribute\":\"\",\"valid\":true,\"id\":\"Qfs4YfuB\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":3}]},\"step_2\":{\"dataSourceStep\":0,\"toolName\":\"polygonTool\",\"result\":[{\"id\":\"ts5jlJxb\",\"sourceID\":\"\",\"valid\":true,\"textAttribute\":\"\",\"pointList\":[{\"x\":178.69738219895285,\"y\":184.93403141361256},{\"x\":204.91937172774865,\"y\":172.5130890052356},{\"x\":252.5329842931937,\"y\":169.0628272251309},{\"x\":295.3162303664921,\"y\":175.27329842931937},{\"x\":333.95916230366487,\"y\":195.2848167539267},{\"x\":360.18115183246067,\"y\":220.1267015706806},{\"x\":376.0523560209424,\"y\":262.909947643979},{\"x\":373.98219895287957,\"y\":296.0324607329843},{\"x\":344.99999999999994,\"y\":335.365445026178},{\"x\":322.22827225130885,\"y\":355.37696335078533},{\"x\":272.544502617801,\"y\":378.1486910994764},{\"x\":221.48062827225127,\"y\":386.42931937172773},{\"x\":187.6680628272251,\"y\":385.7392670157068},{\"x\":158.68586387434553,\"y\":369.1780104712042},{\"x\":137.98429319371724,\"y\":337.43560209424083},{\"x\":127.63350785340312,\"y\":295.34240837696336},{\"x\":131.0837696335078,\"y\":242.89842931937173},{\"x\":147.64502617801045,\"y\":208.3958115183246}],\"attribute\":\"\",\"order\":1}]}}", 11 | "url": "https://user-images.githubusercontent.com/15977946/152742707-c0c51844-e1d0-42d0-9a12-e369002e082f.jpg", 12 | "fileName": "bing_dwen_dwen2.jpg" 13 | }, 14 | { 15 | "id": 3, 16 | "result": "{\"width\":690,\"height\":811,\"valid\":true,\"rotate\":0,\"step_1\":{\"dataSourceStep\":0,\"toolName\":\"pointTool\",\"result\":[{\"x\":361.13507853403144,\"y\":300.62198952879584,\"attribute\":\"\",\"valid\":true,\"id\":\"uAtbXtf2\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":1},{\"x\":242.24502617801048,\"y\":317.60628272251313,\"attribute\":\"\",\"valid\":true,\"id\":\"iLtceHMA\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":2},{\"x\":302.5392670157068,\"y\":356.67015706806285,\"attribute\":\"\",\"valid\":true,\"id\":\"n9MTlJ6A\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":3}]},\"step_2\":{\"dataSourceStep\":0,\"toolName\":\"polygonTool\",\"result\":[{\"id\":\"5sTLU5wF\",\"sourceID\":\"\",\"valid\":true,\"textAttribute\":\"\",\"pointList\":[{\"x\":227.80837696335078,\"y\":247.12146596858642},{\"x\":248.18952879581153,\"y\":235.23246073298432},{\"x\":291.4994764397906,\"y\":225.04188481675394},{\"x\":351.7937172774869,\"y\":229.28795811518327},{\"x\":393.40523560209425,\"y\":245.42303664921468},{\"x\":424.8261780104712,\"y\":272.59790575916236},{\"x\":443.5089005235602,\"y\":298.07434554973827},{\"x\":436.7151832460733,\"y\":345.6303664921466},{\"x\":406.1434554973822,\"y\":382.9958115183247},{\"x\":355.1905759162304,\"y\":408.4722513089006},{\"x\":313.57905759162304,\"y\":419.5120418848168},{\"x\":262.6261780104712,\"y\":417.81361256544506},{\"x\":224.41151832460733,\"y\":399.9801047120419},{\"x\":201.48272251308902,\"y\":364.3130890052356},{\"x\":194.68900523560208,\"y\":315.0586387434555},{\"x\":202.33193717277487,\"y\":272.59790575916236}],\"attribute\":\"\",\"order\":1}]}}", 17 | "url": "https://user-images.githubusercontent.com/15977946/152742728-99392ecf-8f5c-46cf-b5c4-fe7fb6b39976.jpg", 18 | "fileName": "bing_dwen_dwen3.jpg" 19 | }, 20 | { 21 | "id": 4, 22 | "result": "{\"width\":690,\"height\":690,\"valid\":true,\"rotate\":0,\"step_1\":{\"dataSourceStep\":0,\"toolName\":\"pointTool\",\"result\":[{\"x\":365.9528795811519,\"y\":464.5759162303665,\"attribute\":\"\",\"valid\":true,\"id\":\"IKprTuHS\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":1},{\"x\":470.71727748691103,\"y\":445.06806282722516,\"attribute\":\"\",\"valid\":true,\"id\":\"Z90CWkEI\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":2},{\"x\":410.74869109947645,\"y\":395.2146596858639,\"attribute\":\"\",\"valid\":true,\"id\":\"UWRstKZk\",\"sourceID\":\"\",\"textAttribute\":\"\",\"order\":3}]},\"step_2\":{\"dataSourceStep\":0,\"toolName\":\"polygonTool\",\"result\":[{\"id\":\"C30Pc9Ww\",\"sourceID\":\"\",\"valid\":true,\"textAttribute\":\"\",\"pointList\":[{\"x\":412.91623036649213,\"y\":325.85340314136124},{\"x\":468.5497382198953,\"y\":335.9685863874345},{\"x\":501.78534031413614,\"y\":369.2041884816754},{\"x\":514.0680628272252,\"y\":415.44502617801044},{\"x\":504.67539267015707,\"y\":472.5235602094241},{\"x\":484.44502617801044,\"y\":497.0890052356021},{\"x\":443.26178010471205,\"y\":512.9842931937172},{\"x\":389.7958115183246,\"y\":518.7643979057591},{\"x\":336.32984293193715,\"y\":504.31413612565444},{\"x\":302.3717277486911,\"y\":462.40837696335075},{\"x\":298.0366492146597,\"y\":416.89005235602093},{\"x\":318.26701570680626,\"y\":372.0942408376963},{\"x\":363.0628272251309,\"y\":341.0261780104712}],\"attribute\":\"\",\"order\":1}]}}", 23 | "url": "https://user-images.githubusercontent.com/15977946/152742755-9dc75f89-4156-4103-9c6d-f35f1f409d11.jpg", 24 | "fileName": "bing_dwen_dwen4.jpg" 25 | } 26 | ] 27 | -------------------------------------------------------------------------------- /model_configs/mmpose/_base_/datasets/fly.py: -------------------------------------------------------------------------------- 1 | dataset_info = dict( 2 | dataset_name='fly', 3 | paper_info=dict( 4 | author='Pereira, Talmo D and Aldarondo, Diego E and ' 5 | 'Willmore, Lindsay and Kislin, Mikhail and ' 6 | 'Wang, Samuel S-H and Murthy, Mala and Shaevitz, Joshua W', 7 | title='Fast animal pose estimation using deep neural networks', 8 | container='Nature methods', 9 | year='2019', 10 | homepage='https://github.com/jgraving/DeepPoseKit-Data', 11 | ), 12 | keypoint_info={ 13 | 0: 14 | dict(name='head', id=0, color=[255, 255, 255], type='', swap=''), 15 | 1: 16 | dict(name='eyeL', id=1, color=[255, 255, 255], type='', swap='eyeR'), 17 | 2: 18 | dict(name='eyeR', id=2, color=[255, 255, 255], type='', swap='eyeL'), 19 | 3: 20 | dict(name='neck', id=3, color=[255, 255, 255], type='', swap=''), 21 | 4: 22 | dict(name='thorax', id=4, color=[255, 255, 255], type='', swap=''), 23 | 5: 24 | dict(name='abdomen', id=5, color=[255, 255, 255], type='', swap=''), 25 | 6: 26 | dict(name='forelegR1', 27 | id=6, 28 | color=[255, 255, 255], 29 | type='', 30 | swap='forelegL1'), 31 | 7: 32 | dict(name='forelegR2', 33 | id=7, 34 | color=[255, 255, 255], 35 | type='', 36 | swap='forelegL2'), 37 | 8: 38 | dict(name='forelegR3', 39 | id=8, 40 | color=[255, 255, 255], 41 | type='', 42 | swap='forelegL3'), 43 | 9: 44 | dict(name='forelegR4', 45 | id=9, 46 | color=[255, 255, 255], 47 | type='', 48 | swap='forelegL4'), 49 | 10: 50 | dict(name='midlegR1', 51 | id=10, 52 | color=[255, 255, 255], 53 | type='', 54 | swap='midlegL1'), 55 | 11: 56 | dict(name='midlegR2', 57 | id=11, 58 | color=[255, 255, 255], 59 | type='', 60 | swap='midlegL2'), 61 | 12: 62 | dict(name='midlegR3', 63 | id=12, 64 | color=[255, 255, 255], 65 | type='', 66 | swap='midlegL3'), 67 | 13: 68 | dict(name='midlegR4', 69 | id=13, 70 | color=[255, 255, 255], 71 | type='', 72 | swap='midlegL4'), 73 | 14: 74 | dict(name='hindlegR1', 75 | id=14, 76 | color=[255, 255, 255], 77 | type='', 78 | swap='hindlegL1'), 79 | 15: 80 | dict(name='hindlegR2', 81 | id=15, 82 | color=[255, 255, 255], 83 | type='', 84 | swap='hindlegL2'), 85 | 16: 86 | dict(name='hindlegR3', 87 | id=16, 88 | color=[255, 255, 255], 89 | type='', 90 | swap='hindlegL3'), 91 | 17: 92 | dict(name='hindlegR4', 93 | id=17, 94 | color=[255, 255, 255], 95 | type='', 96 | swap='hindlegL4'), 97 | 18: 98 | dict(name='forelegL1', 99 | id=18, 100 | color=[255, 255, 255], 101 | type='', 102 | swap='forelegR1'), 103 | 19: 104 | dict(name='forelegL2', 105 | id=19, 106 | color=[255, 255, 255], 107 | type='', 108 | swap='forelegR2'), 109 | 20: 110 | dict(name='forelegL3', 111 | id=20, 112 | color=[255, 255, 255], 113 | type='', 114 | swap='forelegR3'), 115 | 21: 116 | dict(name='forelegL4', 117 | id=21, 118 | color=[255, 255, 255], 119 | type='', 120 | swap='forelegR4'), 121 | 22: 122 | dict(name='midlegL1', 123 | id=22, 124 | color=[255, 255, 255], 125 | type='', 126 | swap='midlegR1'), 127 | 23: 128 | dict(name='midlegL2', 129 | id=23, 130 | color=[255, 255, 255], 131 | type='', 132 | swap='midlegR2'), 133 | 24: 134 | dict(name='midlegL3', 135 | id=24, 136 | color=[255, 255, 255], 137 | type='', 138 | swap='midlegR3'), 139 | 25: 140 | dict(name='midlegL4', 141 | id=25, 142 | color=[255, 255, 255], 143 | type='', 144 | swap='midlegR4'), 145 | 26: 146 | dict(name='hindlegL1', 147 | id=26, 148 | color=[255, 255, 255], 149 | type='', 150 | swap='hindlegR1'), 151 | 27: 152 | dict(name='hindlegL2', 153 | id=27, 154 | color=[255, 255, 255], 155 | type='', 156 | swap='hindlegR2'), 157 | 28: 158 | dict(name='hindlegL3', 159 | id=28, 160 | color=[255, 255, 255], 161 | type='', 162 | swap='hindlegR3'), 163 | 29: 164 | dict(name='hindlegL4', 165 | id=29, 166 | color=[255, 255, 255], 167 | type='', 168 | swap='hindlegR4'), 169 | 30: 170 | dict(name='wingL', id=30, color=[255, 255, 255], type='', 171 | swap='wingR'), 172 | 31: 173 | dict(name='wingR', id=31, color=[255, 255, 255], type='', 174 | swap='wingL'), 175 | }, 176 | skeleton_info={ 177 | 0: dict(link=('eyeL', 'head'), id=0, color=[255, 255, 255]), 178 | 1: dict(link=('eyeR', 'head'), id=1, color=[255, 255, 255]), 179 | 2: dict(link=('neck', 'head'), id=2, color=[255, 255, 255]), 180 | 3: dict(link=('thorax', 'neck'), id=3, color=[255, 255, 255]), 181 | 4: dict(link=('abdomen', 'thorax'), id=4, color=[255, 255, 255]), 182 | 5: dict(link=('forelegR2', 'forelegR1'), id=5, color=[255, 255, 255]), 183 | 6: dict(link=('forelegR3', 'forelegR2'), id=6, color=[255, 255, 255]), 184 | 7: dict(link=('forelegR4', 'forelegR3'), id=7, color=[255, 255, 255]), 185 | 8: dict(link=('midlegR2', 'midlegR1'), id=8, color=[255, 255, 255]), 186 | 9: dict(link=('midlegR3', 'midlegR2'), id=9, color=[255, 255, 255]), 187 | 10: dict(link=('midlegR4', 'midlegR3'), id=10, color=[255, 255, 255]), 188 | 11: dict(link=('hindlegR2', 'hindlegR1'), id=11, color=[255, 255, 189 | 255]), 190 | 12: dict(link=('hindlegR3', 'hindlegR2'), id=12, color=[255, 255, 191 | 255]), 192 | 13: dict(link=('hindlegR4', 'hindlegR3'), id=13, color=[255, 255, 193 | 255]), 194 | 14: dict(link=('forelegL2', 'forelegL1'), id=14, color=[255, 255, 195 | 255]), 196 | 15: dict(link=('forelegL3', 'forelegL2'), id=15, color=[255, 255, 197 | 255]), 198 | 16: dict(link=('forelegL4', 'forelegL3'), id=16, color=[255, 255, 199 | 255]), 200 | 17: dict(link=('midlegL2', 'midlegL1'), id=17, color=[255, 255, 255]), 201 | 18: dict(link=('midlegL3', 'midlegL2'), id=18, color=[255, 255, 255]), 202 | 19: dict(link=('midlegL4', 'midlegL3'), id=19, color=[255, 255, 255]), 203 | 20: dict(link=('hindlegL2', 'hindlegL1'), id=20, color=[255, 255, 204 | 255]), 205 | 21: dict(link=('hindlegL3', 'hindlegL2'), id=21, color=[255, 255, 206 | 255]), 207 | 22: dict(link=('hindlegL4', 'hindlegL3'), id=22, color=[255, 255, 208 | 255]), 209 | 23: dict(link=('wingL', 'neck'), id=23, color=[255, 255, 255]), 210 | 24: dict(link=('wingR', 'neck'), id=24, color=[255, 255, 255]) 211 | }, 212 | joint_weights=[1.] * 32, 213 | sigmas=[]) 214 | --------------------------------------------------------------------------------