├── .gitignore
├── HRFormer-20-fps.gif
├── LICENSE
├── README.md
├── cls
├── .gitattributes
├── .gitignore
├── LICENSE
├── README.md
├── config.py
├── configs
│ ├── deit
│ │ ├── deit_base_patch16_224.yaml
│ │ ├── deit_small_patch16_224.yaml
│ │ └── deit_tiny_patch16_224.yaml
│ ├── hrnet
│ │ ├── hrnet_w18.yaml
│ │ ├── hrnet_w18_small_v2.yaml
│ │ ├── hrnet_w32.yaml
│ │ └── hrnet_w48.yaml
│ ├── hrt
│ │ ├── hrt_base.yaml
│ │ ├── hrt_small.yaml
│ │ └── hrt_tiny.yaml
│ ├── pvt
│ │ ├── pvt_v2_b0_patch16_224.yaml
│ │ ├── pvt_v2_b1_patch16_224.yaml
│ │ ├── pvt_v2_b2_patch16_224.yaml
│ │ ├── pvt_v2_b3_patch16_224.yaml
│ │ ├── pvt_v2_b4_patch16_224.yaml
│ │ └── pvt_v2_b5_patch16_224.yaml
│ └── swin
│ │ ├── swin_base_patch4_window12_384.yaml
│ │ ├── swin_base_patch4_window7_224.yaml
│ │ ├── swin_large_patch4_window12_384.yaml
│ │ ├── swin_large_patch4_window7_224.yaml
│ │ ├── swin_small_patch4_window7_224.yaml
│ │ └── swin_tiny_patch4_window7_224.yaml
├── data
│ ├── __init__.py
│ ├── build.py
│ ├── cached_image_folder.py
│ ├── samplers.py
│ └── zipreader.py
├── figures
│ ├── HRFormer.png
│ ├── HRFormerUnit.png
│ ├── HRT_arch5.pdf
│ ├── HRT_arch5.png
│ ├── HRT_layer6.pdf
│ └── HRT_layer6.png
├── install.sh
├── logger.py
├── lr_scheduler.py
├── main.py
├── models
│ ├── __init__.py
│ ├── build.py
│ ├── hrnet.py
│ ├── hrt.py
│ ├── modules
│ │ ├── __init__.py
│ │ ├── basic_block.py
│ │ ├── bottleneck_block.py
│ │ ├── ffn_block.py
│ │ ├── multihead_attention.py
│ │ ├── multihead_isa_attention.py
│ │ └── transformer_block.py
│ ├── pvt_v2.py
│ ├── swin_transformer.py
│ └── vision_transformer.py
├── optimizer.py
├── run_dist.sh
├── run_eval.sh
├── run_flops.sh
├── test_flops.py
├── tools
│ ├── __init__.py
│ ├── flop_count.py
│ └── jit_handles.py
└── utils.py
├── pose
├── .github
│ ├── CODE_OF_CONDUCT.md
│ ├── CONTRIBUTING.md
│ ├── ISSUE_TEMPLATE
│ │ ├── config.yml
│ │ ├── error-report.md
│ │ ├── feature_request.md
│ │ ├── general_questions.md
│ │ └── reimplementation_questions.md
│ └── workflows
│ │ ├── build.yml
│ │ └── deploy.yml
├── .gitignore
├── .pre-commit-config.yaml
├── .pylintrc
├── .readthedocs.yml
├── LICENSE
├── LOG
├── README.md
├── README_CN.md
├── README_MMPOSE.md
├── configs
│ ├── animal
│ │ ├── README.md
│ │ ├── hrnet
│ │ │ ├── README.md
│ │ │ ├── atrw
│ │ │ │ ├── hrnet_w32_atrw_256x256.py
│ │ │ │ └── hrnet_w48_atrw_256x256.py
│ │ │ ├── horse10
│ │ │ │ ├── hrnet_w32_horse10_256x256-split1.py
│ │ │ │ ├── hrnet_w32_horse10_256x256-split2.py
│ │ │ │ ├── hrnet_w32_horse10_256x256-split3.py
│ │ │ │ ├── hrnet_w48_horse10_256x256-split1.py
│ │ │ │ ├── hrnet_w48_horse10_256x256-split2.py
│ │ │ │ └── hrnet_w48_horse10_256x256-split3.py
│ │ │ └── macaque
│ │ │ │ ├── hrnet_w32_macaque_256x192.py
│ │ │ │ └── hrnet_w48_macaque_256x192.py
│ │ └── resnet
│ │ │ ├── README.md
│ │ │ ├── atrw
│ │ │ ├── res101_atrw_256x256.py
│ │ │ ├── res152_atrw_256x256.py
│ │ │ └── res50_atrw_256x256.py
│ │ │ ├── fly
│ │ │ ├── res101_fly_192x192.py
│ │ │ ├── res152_fly_192x192.py
│ │ │ └── res50_fly_192x192.py
│ │ │ ├── horse10
│ │ │ ├── res101_horse10_256x256-split1.py
│ │ │ ├── res101_horse10_256x256-split2.py
│ │ │ ├── res101_horse10_256x256-split3.py
│ │ │ ├── res152_horse10_256x256-split1.py
│ │ │ ├── res152_horse10_256x256-split2.py
│ │ │ ├── res152_horse10_256x256-split3.py
│ │ │ ├── res50_horse10_256x256-split1.py
│ │ │ ├── res50_horse10_256x256-split2.py
│ │ │ └── res50_horse10_256x256-split3.py
│ │ │ ├── locust
│ │ │ ├── res101_locust_160x160.py
│ │ │ ├── res152_locust_160x160.py
│ │ │ └── res50_locust_160x160.py
│ │ │ ├── macaque
│ │ │ ├── res101_macaque_256x192.py
│ │ │ ├── res152_macaque_256x192.py
│ │ │ └── res50_macaque_256x192.py
│ │ │ └── zebra
│ │ │ ├── res101_zebra_160x160.py
│ │ │ ├── res152_zebra_160x160.py
│ │ │ └── res50_zebra_160x160.py
│ ├── body3d
│ │ └── simple_baseline
│ │ │ ├── README.md
│ │ │ └── h36m
│ │ │ └── simple3Dbaseline_h36m.py
│ ├── bottom_up
│ │ ├── README.md
│ │ ├── higherhrnet
│ │ │ ├── README.md
│ │ │ ├── aic
│ │ │ │ └── higher_hrnet32_aic_512x512.py
│ │ │ ├── coco
│ │ │ │ ├── higher_hrnet32_coco_512x512.py
│ │ │ │ ├── higher_hrnet32_coco_640x640.py
│ │ │ │ └── higher_hrnet48_coco_512x512.py
│ │ │ └── crowdpose
│ │ │ │ ├── higher_hrnet32_crowdpose_512x512.py
│ │ │ │ ├── higher_hrnet32_crowdpose_640x640.py
│ │ │ │ └── higher_hrnet48_crowdpose_512x512.py
│ │ ├── hrnet
│ │ │ ├── README.md
│ │ │ ├── aic
│ │ │ │ └── hrnet_w32_aic_512x512.py
│ │ │ ├── coco
│ │ │ │ ├── hrnet_w32_coco_512x512.py
│ │ │ │ ├── hrnet_w32_coco_640x640.py
│ │ │ │ ├── hrnet_w48_coco_512x512.py
│ │ │ │ └── hrnet_w48_coco_640x640.py
│ │ │ └── mhp
│ │ │ │ └── hrnet_w48_mhp_512x512.py
│ │ ├── mobilenet
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ └── mobilenetv2_coco_512x512.py
│ │ │ └── crowdpose
│ │ │ │ └── mobilenetv2_crowdpose_512x512.py
│ │ ├── resnet
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── res101_coco_512x512.py
│ │ │ │ ├── res101_coco_640x640.py
│ │ │ │ ├── res152_coco_512x512.py
│ │ │ │ ├── res152_coco_640x640.py
│ │ │ │ ├── res50_coco_512x512.py
│ │ │ │ └── res50_coco_640x640.py
│ │ │ └── crowdpose
│ │ │ │ ├── res101_crowdpose_512x512.py
│ │ │ │ ├── res152_crowdpose_512x512.py
│ │ │ │ └── res50_crowdpose_512x512.py
│ │ └── udp
│ │ │ ├── README.md
│ │ │ ├── aic
│ │ │ └── higher_hrnet32_aic_512x512_udp.py
│ │ │ ├── coco
│ │ │ ├── higher_hrnet32_coco_512x512_udp.py
│ │ │ ├── higher_hrnet32_coco_640x640_udp.py
│ │ │ ├── higher_hrnet48_coco_512x512_udp.py
│ │ │ ├── hrnet_w32_coco_512x512_udp.py
│ │ │ ├── hrnet_w32_coco_640x640_udp.py
│ │ │ ├── hrnet_w48_coco_512x512_udp.py
│ │ │ └── hrnet_w48_coco_640x640_udp.py
│ │ │ └── crowdpose
│ │ │ ├── higher_hrnet32_crowdpose_512x512_udp.py
│ │ │ ├── higher_hrnet32_crowdpose_640x640_udp.py
│ │ │ └── higher_hrnet48_crowdpose_512x512_udp.py
│ ├── face
│ │ ├── README.md
│ │ ├── darkpose
│ │ │ ├── 300w
│ │ │ │ └── hrnetv2_w18_300w_256x256_dark.py
│ │ │ ├── README.md
│ │ │ ├── aflw
│ │ │ │ └── hrnetv2_w18_aflw_256x256_dark.py
│ │ │ ├── cofw
│ │ │ │ └── hrnetv2_w18_cofw_256x256_dark.py
│ │ │ └── wflw
│ │ │ │ └── hrnetv2_w18_wflw_256x256_dark.py
│ │ ├── deeppose
│ │ │ ├── README.md
│ │ │ └── wflw
│ │ │ │ ├── deeppose_res50_wflw_256x256.py
│ │ │ │ └── deeppose_res50_wflw_256x256_wingloss.py
│ │ ├── hrnetv2
│ │ │ ├── 300w
│ │ │ │ └── hrnetv2_w18_300w_256x256.py
│ │ │ ├── README.md
│ │ │ ├── aflw
│ │ │ │ └── hrnetv2_w18_aflw_256x256.py
│ │ │ ├── cofw
│ │ │ │ └── hrnetv2_w18_cofw_256x256.py
│ │ │ └── wflw
│ │ │ │ └── hrnetv2_w18_wflw_256x256.py
│ │ └── resnet
│ │ │ ├── 300w
│ │ │ └── res50_300w_256x256.py
│ │ │ ├── README.md
│ │ │ ├── aflw
│ │ │ └── res50_aflw_256x256.py
│ │ │ ├── cofw
│ │ │ └── res50_cofw_256x256.py
│ │ │ └── wflw
│ │ │ └── res50_wflw_256x256.py
│ ├── fashion
│ │ ├── README.md
│ │ ├── deeppose
│ │ │ ├── README.md
│ │ │ └── deepfashion
│ │ │ │ ├── deeppose_res101_deepfashion_full_256x192.py
│ │ │ │ ├── deeppose_res101_deepfashion_lower_256x192.py
│ │ │ │ ├── deeppose_res101_deepfashion_upper_256x192.py
│ │ │ │ ├── deeppose_res152_deepfashion_full_256x192.py
│ │ │ │ ├── deeppose_res152_deepfashion_lower_256x192.py
│ │ │ │ ├── deeppose_res152_deepfashion_upper_256x192.py
│ │ │ │ ├── deeppose_res50_deepfashion_full_256x192.py
│ │ │ │ ├── deeppose_res50_deepfashion_lower_256x192.py
│ │ │ │ └── deeppose_res50_deepfashion_upper_256x192.py
│ │ ├── hrnet
│ │ │ └── deepfashion
│ │ │ │ ├── hrnet_w32_deepfashion_full_256x192.py
│ │ │ │ ├── hrnet_w32_deepfashion_lower_256x192.py
│ │ │ │ ├── hrnet_w32_deepfashion_upper_256x192.py
│ │ │ │ ├── hrnet_w48_deepfashion_full_256x192.py
│ │ │ │ ├── hrnet_w48_deepfashion_lower_256x192.py
│ │ │ │ └── hrnet_w48_deepfashion_upper_256x192.py
│ │ ├── resnet
│ │ │ ├── README.md
│ │ │ └── deepfashion
│ │ │ │ ├── res101_deepfashion_full_256x192.py
│ │ │ │ ├── res101_deepfashion_lower_256x192.py
│ │ │ │ ├── res101_deepfashion_upper_256x192.py
│ │ │ │ ├── res152_deepfashion_full_256x192.py
│ │ │ │ ├── res152_deepfashion_lower_256x192.py
│ │ │ │ ├── res152_deepfashion_upper_256x192.py
│ │ │ │ ├── res50_deepfashion_full_256x192.py
│ │ │ │ ├── res50_deepfashion_lower_256x192.py
│ │ │ │ └── res50_deepfashion_upper_256x192.py
│ │ └── udp
│ │ │ └── deepfashion
│ │ │ ├── hrnet_w32_deepfashion_full_256x192_udp.py
│ │ │ ├── hrnet_w32_deepfashion_lower_256x192_udp.py
│ │ │ ├── hrnet_w32_deepfashion_upper_256x192_udp.py
│ │ │ ├── hrnet_w48_deepfashion_full_256x192_udp.py
│ │ │ ├── hrnet_w48_deepfashion_lower_256x192_udp.py
│ │ │ └── hrnet_w48_deepfashion_upper_256x192_udp.py
│ ├── hand
│ │ ├── README.md
│ │ ├── darkpose
│ │ │ ├── README.md
│ │ │ ├── onehand10k
│ │ │ │ └── hrnetv2_w18_onehand10k_256x256_dark.py
│ │ │ ├── panoptic
│ │ │ │ └── hrnetv2_w18_panoptic_256x256_dark.py
│ │ │ └── rhd2d
│ │ │ │ └── hrnetv2_w18_rhd2d_256x256_dark.py
│ │ ├── deeppose
│ │ │ ├── README.md
│ │ │ ├── onehand10k
│ │ │ │ └── deeppose_res50_onehand10k_256x256.py
│ │ │ ├── panoptic
│ │ │ │ └── deeppose_res50_panoptic_256x256.py
│ │ │ └── rhd2d
│ │ │ │ └── deeppose_res50_rhd2d_256x256.py
│ │ ├── hrnetv2
│ │ │ ├── README.md
│ │ │ ├── onehand10k
│ │ │ │ └── hrnetv2_w18_onehand10k_256x256.py
│ │ │ ├── panoptic
│ │ │ │ └── hrnetv2_w18_panoptic_256x256.py
│ │ │ └── rhd2d
│ │ │ │ └── hrnetv2_w18_rhd2d_256x256.py
│ │ ├── mobilenet_v2
│ │ │ ├── README.md
│ │ │ ├── onehand10k
│ │ │ │ └── mobilenetv2_onehand10k_256x256.py
│ │ │ ├── panoptic
│ │ │ │ └── mobilenetv2_panoptic_256x256.py
│ │ │ └── rhd2d
│ │ │ │ └── mobilenetv2_rhd2d_256x256.py
│ │ ├── resnet
│ │ │ ├── README.md
│ │ │ ├── freihand
│ │ │ │ └── res50_freihand_224x224.py
│ │ │ ├── interhand2d
│ │ │ │ ├── res50_interhand2d_all_256x256.py
│ │ │ │ ├── res50_interhand2d_human_256x256.py
│ │ │ │ └── res50_interhand2d_machine_256x256.py
│ │ │ ├── onehand10k
│ │ │ │ └── res50_onehand10k_256x256.py
│ │ │ ├── panoptic
│ │ │ │ └── res50_panoptic_256x256.py
│ │ │ └── rhd2d
│ │ │ │ ├── res50_rhd2d_224x224.py
│ │ │ │ └── res50_rhd2d_256x256.py
│ │ └── udp
│ │ │ ├── README.md
│ │ │ ├── onehand10k
│ │ │ └── hrnetv2_w18_onehand10k_256x256_udp.py
│ │ │ ├── panoptic
│ │ │ └── hrnetv2_w18_panoptic_256x256_udp.py
│ │ │ └── rhd2d
│ │ │ └── hrnetv2_w18_rhd2d_256x256_udp.py
│ ├── mesh
│ │ ├── README.md
│ │ └── hmr
│ │ │ ├── README.md
│ │ │ └── hmr_res50_224x224.py
│ ├── top_down
│ │ ├── README.md
│ │ ├── alexnet
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ └── alexnet_coco_256x192.py
│ │ ├── augmentation
│ │ │ ├── README.md
│ │ │ ├── hrnet_w32_coco_256x192_coarsedropout.py
│ │ │ ├── hrnet_w32_coco_256x192_gridmask.py
│ │ │ └── hrnet_w32_coco_256x192_photometric.py
│ │ ├── cpm
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── cpm_coco_256x192.py
│ │ │ │ └── cpm_coco_384x288.py
│ │ │ ├── jhmdb
│ │ │ │ ├── cpm_jhmdb_sub1_368x368.py
│ │ │ │ ├── cpm_jhmdb_sub2_368x368.py
│ │ │ │ └── cpm_jhmdb_sub3_368x368.py
│ │ │ └── mpii
│ │ │ │ └── cpm_mpii_368x368.py
│ │ ├── darkpose
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── hrnet_w32_coco_256x192_dark.py
│ │ │ │ ├── hrnet_w32_coco_384x288_dark.py
│ │ │ │ ├── hrnet_w48_coco_256x192_dark.py
│ │ │ │ ├── hrnet_w48_coco_384x288_dark.py
│ │ │ │ ├── res101_coco_256x192_dark.py
│ │ │ │ ├── res101_coco_384x288_dark.py
│ │ │ │ ├── res152_coco_256x192_dark.py
│ │ │ │ ├── res152_coco_384x288_dark.py
│ │ │ │ ├── res50_coco_256x192_dark.py
│ │ │ │ └── res50_coco_384x288_dark.py
│ │ │ └── mpii
│ │ │ │ ├── hrnet_w32_mpii_256x256_dark.py
│ │ │ │ └── hrnet_w48_mpii_256x256_dark.py
│ │ ├── deeppose
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── deeppose_res101_coco_256x192.py
│ │ │ │ ├── deeppose_res152_coco_256x192.py
│ │ │ │ └── deeppose_res50_coco_256x192.py
│ │ │ └── mpii
│ │ │ │ ├── deeppose_res101_mpii_256x256.py
│ │ │ │ ├── deeppose_res152_mpii_256x256.py
│ │ │ │ └── deeppose_res50_mpii_256x256.py
│ │ ├── hourglass
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── hourglass52_coco_256x256.py
│ │ │ │ └── hourglass52_coco_384x384.py
│ │ │ └── mpii
│ │ │ │ ├── hourglass52_mpii_256x256.py
│ │ │ │ └── hourglass52_mpii_384x384.py
│ │ ├── hrnet
│ │ │ ├── README.md
│ │ │ ├── aic
│ │ │ │ ├── hrnet_w32_aic_256x192.py
│ │ │ │ ├── hrnet_w32_aic_384x288.py
│ │ │ │ ├── hrnet_w48_aic_256x192.py
│ │ │ │ └── hrnet_w48_aic_384x288.py
│ │ │ ├── coco
│ │ │ │ ├── hrnet_w18_coco_256x192.py
│ │ │ │ ├── hrnet_w32_coco_256x192.py
│ │ │ │ ├── hrnet_w32_coco_384x288.py
│ │ │ │ ├── hrnet_w48_coco_256x192.py
│ │ │ │ └── hrnet_w48_coco_384x288.py
│ │ │ ├── crowdpose
│ │ │ │ ├── hrnet_w32_crowdpose_256x192.py
│ │ │ │ ├── hrnet_w32_crowdpose_384x288.py
│ │ │ │ ├── hrnet_w48_crowdpose_256x192.py
│ │ │ │ └── hrnet_w48_crowdpose_384x288.py
│ │ │ ├── mpii
│ │ │ │ ├── hrnet_w32_mpii_256x256.py
│ │ │ │ └── hrnet_w48_mpii_256x256.py
│ │ │ └── posetrack18
│ │ │ │ └── hrnet_w32_posetrack18_256x192.py
│ │ ├── hrt
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ ├── hrt_base_coco_256x192.py
│ │ │ │ ├── hrt_base_coco_384x288.py
│ │ │ │ ├── hrt_small_coco_256x192.py
│ │ │ │ └── hrt_small_coco_384x288.py
│ │ ├── mobilenet_v2
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── mobilenetv2_coco_256x192.py
│ │ │ │ └── mobilenetv2_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ └── mobilenetv2_mpii_256x256.py
│ │ ├── mspn
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ ├── 2xmspn50_coco_256x192.py
│ │ │ │ ├── 3xmspn50_coco_256x192.py
│ │ │ │ ├── 4xmspn50_coco_256x192.py
│ │ │ │ └── mspn50_coco_256x192.py
│ │ ├── resnest
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ ├── resnest101_coco_256x192.py
│ │ │ │ ├── resnest101_coco_384x288.py
│ │ │ │ ├── resnest50_coco_256x192.py
│ │ │ │ └── resnest50_coco_384x288.py
│ │ ├── resnet
│ │ │ ├── README.md
│ │ │ ├── aic
│ │ │ │ ├── res101_aic_256x192.py
│ │ │ │ ├── res101_aic_384x288.py
│ │ │ │ ├── res152_aic_256x192.py
│ │ │ │ ├── res152_aic_384x288.py
│ │ │ │ ├── res50_aic_256x192.py
│ │ │ │ └── res50_aic_384x288.py
│ │ │ ├── coco
│ │ │ │ ├── res101_coco_256x192.py
│ │ │ │ ├── res101_coco_384x288.py
│ │ │ │ ├── res152_coco_256x192.py
│ │ │ │ ├── res152_coco_384x288.py
│ │ │ │ ├── res50_coco_256x192.py
│ │ │ │ └── res50_coco_384x288.py
│ │ │ ├── crowdpose
│ │ │ │ ├── res101_crowdpose_256x192.py
│ │ │ │ ├── res101_crowdpose_320x256.py
│ │ │ │ ├── res101_crowdpose_384x288.py
│ │ │ │ ├── res152_crowdpose_256x192.py
│ │ │ │ ├── res152_crowdpose_384x288.py
│ │ │ │ ├── res50_crowdpose_256x192.py
│ │ │ │ └── res50_crowdpose_384x288.py
│ │ │ ├── jhmdb
│ │ │ │ ├── res50_2deconv_jhmdb_sub1_256x256.py
│ │ │ │ ├── res50_2deconv_jhmdb_sub2_256x256.py
│ │ │ │ ├── res50_2deconv_jhmdb_sub3_256x256.py
│ │ │ │ ├── res50_jhmdb_sub1_256x256.py
│ │ │ │ ├── res50_jhmdb_sub2_256x256.py
│ │ │ │ └── res50_jhmdb_sub3_256x256.py
│ │ │ ├── mhp
│ │ │ │ └── res50_mhp_256x192.py
│ │ │ ├── mpii
│ │ │ │ ├── res101_mpii_256x256.py
│ │ │ │ ├── res152_mpii_256x256.py
│ │ │ │ └── res50_mpii_256x256.py
│ │ │ ├── mpii_trb
│ │ │ │ ├── res101_mpii_trb_256x256.py
│ │ │ │ ├── res152_mpii_trb_256x256.py
│ │ │ │ └── res50_mpii_trb_256x256.py
│ │ │ ├── ochuman
│ │ │ │ ├── res101_ochuman_256x192.py
│ │ │ │ ├── res101_ochuman_384x288.py
│ │ │ │ ├── res152_ochuman_256x192.py
│ │ │ │ ├── res152_ochuman_384x288.py
│ │ │ │ ├── res50_ochuman_256x192.py
│ │ │ │ └── res50_ochuman_384x288.py
│ │ │ └── posetrack18
│ │ │ │ └── res50_posetrack18_256x192.py
│ │ ├── resnetv1d
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── resnetv1d101_coco_256x192.py
│ │ │ │ ├── resnetv1d101_coco_384x288.py
│ │ │ │ ├── resnetv1d152_coco_256x192.py
│ │ │ │ ├── resnetv1d152_coco_384x288.py
│ │ │ │ ├── resnetv1d50_coco_256x192.py
│ │ │ │ └── resnetv1d50_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ ├── resnetv1d101_mpii_256x256.py
│ │ │ │ ├── resnetv1d152_mpii_256x256.py
│ │ │ │ └── resnetv1d50_mpii_256x256.py
│ │ ├── resnext
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── resnext101_coco_256x192.py
│ │ │ │ ├── resnext101_coco_384x288.py
│ │ │ │ ├── resnext152_coco_256x192.py
│ │ │ │ ├── resnext152_coco_384x288.py
│ │ │ │ ├── resnext50_coco_256x192.py
│ │ │ │ └── resnext50_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ ├── resnext101_mpii_256x256.py
│ │ │ │ ├── resnext152_mpii_256x256.py
│ │ │ │ └── resnext50_mpii_256x256.py
│ │ ├── rsn
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ ├── 2xrsn50_coco_256x192.py
│ │ │ │ ├── 3xrsn50_coco_256x192.py
│ │ │ │ ├── rsn18_coco_256x192.py
│ │ │ │ └── rsn50_coco_256x192.py
│ │ ├── scnet
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── scnet101_coco_256x192.py
│ │ │ │ ├── scnet101_coco_384x288.py
│ │ │ │ ├── scnet50_coco_256x192.py
│ │ │ │ └── scnet50_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ ├── scnet101_mpii_256x256.py
│ │ │ │ └── scnet50_mpii_256x256.py
│ │ ├── seresnet
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── seresnet101_coco_256x192.py
│ │ │ │ ├── seresnet101_coco_384x288.py
│ │ │ │ ├── seresnet152_coco_256x192.py
│ │ │ │ ├── seresnet152_coco_384x288.py
│ │ │ │ ├── seresnet50_coco_256x192.py
│ │ │ │ └── seresnet50_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ ├── seresnet101_mpii_256x256.py
│ │ │ │ ├── seresnet152_mpii_256x256.py
│ │ │ │ └── seresnet50_mpii_256x256.py
│ │ ├── shufflenet_v1
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── shufflenetv1_coco_256x192.py
│ │ │ │ └── shufflenetv1_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ └── shufflenetv1_mpii_256x256.py
│ │ ├── shufflenet_v2
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── shufflenetv2_coco_256x192.py
│ │ │ │ └── shufflenetv2_coco_384x288.py
│ │ │ └── mpii
│ │ │ │ └── shufflenetv2_mpii_256x256.py
│ │ ├── udp
│ │ │ ├── README.md
│ │ │ ├── coco
│ │ │ │ ├── hrnet_w32_coco_256x192_udp.py
│ │ │ │ ├── hrnet_w32_coco_256x192_udp_regress.py
│ │ │ │ ├── hrnet_w32_coco_384x288_udp.py
│ │ │ │ ├── hrnet_w48_coco_256x192_udp.py
│ │ │ │ └── hrnet_w48_coco_384x288_udp.py
│ │ │ └── mpii
│ │ │ │ ├── hrnet_w32_mpii_256x256_udp.py
│ │ │ │ └── hrnet_w48_mpii_256x256_udp.py
│ │ ├── vgg
│ │ │ ├── README.md
│ │ │ └── coco
│ │ │ │ └── vgg16_bn_coco_256x192.py
│ │ └── vit
│ │ │ └── coco
│ │ │ ├── deit_base_distilled_patch16_384_coco_256x192.py
│ │ │ ├── deit_base_distilled_patch8_384_coco_256x192.py
│ │ │ ├── vit_large_patch16_384_coco_256x192.py
│ │ │ └── vit_large_patch8_384_coco_256x192.py
│ └── wholebody
│ │ ├── README.md
│ │ ├── darkpose
│ │ ├── README.md
│ │ └── coco-wholebody
│ │ │ ├── hrnet_w32_coco_wholebody_256x192_dark.py
│ │ │ ├── hrnet_w32_coco_wholebody_384x288_dark.py
│ │ │ ├── hrnet_w48_coco_wholebody_256x192_dark.py
│ │ │ ├── hrnet_w48_coco_wholebody_384x288_dark.py
│ │ │ └── hrnet_w48_coco_wholebody_384x288_dark_plus.py
│ │ ├── hrnet
│ │ ├── README.md
│ │ └── coco-wholebody
│ │ │ ├── hrnet_w32_coco_wholebody_256x192.py
│ │ │ ├── hrnet_w32_coco_wholebody_384x288.py
│ │ │ ├── hrnet_w48_coco_wholebody_256x192.py
│ │ │ └── hrnet_w48_coco_wholebody_384x288.py
│ │ └── resnet
│ │ ├── README.md
│ │ └── coco-wholebody
│ │ ├── res101_coco_wholebody_256x192.py
│ │ ├── res101_coco_wholebody_384x288.py
│ │ ├── res152_coco_wholebody_256x192.py
│ │ ├── res152_coco_wholebody_384x288.py
│ │ ├── res50_coco_wholebody_256x192.py
│ │ └── res50_coco_wholebody_384x288.py
├── demo
│ ├── README.md
│ ├── bottom_up_img_demo.py
│ ├── bottom_up_video_demo.py
│ ├── docs
│ │ ├── 2d_animal_demo.md
│ │ ├── 2d_face_demo.md
│ │ ├── 2d_hand_demo.md
│ │ ├── 2d_human_pose_demo.md
│ │ ├── 2d_pose_tracking_demo.md
│ │ ├── 2d_wholebody_pose_demo.md
│ │ └── mmdet_modelzoo.md
│ ├── face_img_demo.py
│ ├── face_video_demo.py
│ ├── mmdetection_cfg
│ │ ├── cascade_rcnn_x101_64x4d_fpn_1class.py
│ │ ├── cascade_rcnn_x101_64x4d_fpn_coco.py
│ │ ├── faster_rcnn_r50_fpn_1class.py
│ │ └── faster_rcnn_r50_fpn_coco.py
│ ├── mmtracking_cfg
│ │ └── tracktor_faster-rcnn_r50_fpn_4e_mot17-private.py
│ ├── resources
│ │ ├── HRT_arch5.png
│ │ ├── demo.mp4
│ │ └── demo_coco.gif
│ ├── top_down_img_demo.py
│ ├── top_down_img_demo_with_mmdet.py
│ ├── top_down_pose_tracking_demo_with_mmdet.py
│ ├── top_down_pose_tracking_demo_with_mmtracking.py
│ ├── top_down_video_demo_full_frame_without_det.py
│ └── top_down_video_demo_with_mmdet.py
├── docker
│ └── Dockerfile
├── docs
│ ├── Makefile
│ ├── api.rst
│ ├── benchmark.md
│ ├── changelog.md
│ ├── conf.py
│ ├── data_preparation.md
│ ├── faq.md
│ ├── getting_started.md
│ ├── imgs
│ │ ├── acc_curve.png
│ │ ├── qq_group_qrcode.jpg
│ │ └── zhihu_qrcode.jpg
│ ├── index.rst
│ ├── install.md
│ ├── make.bat
│ ├── merge_docs.sh
│ ├── stats.py
│ ├── tasks
│ │ ├── 2d_animal_keypoint.md
│ │ ├── 2d_body_keypoint.md
│ │ ├── 2d_face_keypoint.md
│ │ ├── 2d_fashion_landmark.md
│ │ ├── 2d_hand_keypoint.md
│ │ ├── 2d_wholebody_keypoint.md
│ │ ├── 3d_body_keypoint.md
│ │ └── 3d_body_mesh.md
│ ├── tutorials
│ │ ├── 0_config.md
│ │ ├── 1_finetune.md
│ │ ├── 2_new_dataset.md
│ │ ├── 3_data_pipeline.md
│ │ ├── 4_new_modules.md
│ │ ├── 5_export_model.md
│ │ └── 6_customize_runtime.md
│ └── useful_tools.md
├── mmcv_custom
│ ├── __init__.py
│ └── checkpoint.py
├── mmpose
│ ├── __init__.py
│ ├── apis
│ │ ├── __init__.py
│ │ ├── inference.py
│ │ ├── inference_tracking.py
│ │ ├── test.py
│ │ └── train.py
│ ├── core
│ │ ├── __init__.py
│ │ ├── camera
│ │ │ ├── __init__.py
│ │ │ ├── camera_base.py
│ │ │ └── single_camera.py
│ │ ├── distributed_wrapper.py
│ │ ├── evaluation
│ │ │ ├── __init__.py
│ │ │ ├── bottom_up_eval.py
│ │ │ ├── eval_hooks.py
│ │ │ ├── mesh_eval.py
│ │ │ ├── pose3d_eval.py
│ │ │ └── top_down_eval.py
│ │ ├── fp16
│ │ │ ├── __init__.py
│ │ │ ├── decorators.py
│ │ │ ├── hooks.py
│ │ │ └── utils.py
│ │ ├── optimizer
│ │ │ ├── __init__.py
│ │ │ ├── builder.py
│ │ │ └── registry.py
│ │ ├── post_processing
│ │ │ ├── __init__.py
│ │ │ ├── group.py
│ │ │ ├── nms.py
│ │ │ └── post_transforms.py
│ │ └── utils
│ │ │ ├── __init__.py
│ │ │ ├── dist_utils.py
│ │ │ └── regularizations.py
│ ├── datasets
│ │ ├── __init__.py
│ │ ├── builder.py
│ │ ├── dataset_wrappers.py
│ │ ├── datasets
│ │ │ ├── __init__.py
│ │ │ ├── animal
│ │ │ │ ├── __init__.py
│ │ │ │ ├── animal_atrw_dataset.py
│ │ │ │ ├── animal_base_dataset.py
│ │ │ │ ├── animal_fly_dataset.py
│ │ │ │ ├── animal_horse10_dataset.py
│ │ │ │ ├── animal_locust_dataset.py
│ │ │ │ ├── animal_macaque_dataset.py
│ │ │ │ └── animal_zebra_dataset.py
│ │ │ ├── body3d
│ │ │ │ ├── __init__.py
│ │ │ │ ├── body3d_base_dataset.py
│ │ │ │ └── body3d_h36m_dataset.py
│ │ │ ├── bottom_up
│ │ │ │ ├── __init__.py
│ │ │ │ ├── bottom_up_aic.py
│ │ │ │ ├── bottom_up_base_dataset.py
│ │ │ │ ├── bottom_up_coco.py
│ │ │ │ ├── bottom_up_crowdpose.py
│ │ │ │ └── bottom_up_mhp.py
│ │ │ ├── face
│ │ │ │ ├── __init__.py
│ │ │ │ ├── face_300w_dataset.py
│ │ │ │ ├── face_aflw_dataset.py
│ │ │ │ ├── face_base_dataset.py
│ │ │ │ ├── face_cofw_dataset.py
│ │ │ │ └── face_wflw_dataset.py
│ │ │ ├── fashion
│ │ │ │ ├── __init__.py
│ │ │ │ ├── deepfashion_dataset.py
│ │ │ │ └── fashion_base_dataset.py
│ │ │ ├── hand
│ │ │ │ ├── __init__.py
│ │ │ │ ├── freihand_dataset.py
│ │ │ │ ├── hand_base_dataset.py
│ │ │ │ ├── interhand2d_dataset.py
│ │ │ │ ├── interhand3d_dataset.py
│ │ │ │ ├── onehand10k_dataset.py
│ │ │ │ ├── panoptic_dataset.py
│ │ │ │ └── rhd2d_dataset.py
│ │ │ ├── mesh
│ │ │ │ ├── __init__.py
│ │ │ │ ├── mesh_adv_dataset.py
│ │ │ │ ├── mesh_base_dataset.py
│ │ │ │ ├── mesh_h36m_dataset.py
│ │ │ │ ├── mesh_mix_dataset.py
│ │ │ │ └── mosh_dataset.py
│ │ │ └── top_down
│ │ │ │ ├── __init__.py
│ │ │ │ ├── topdown_aic_dataset.py
│ │ │ │ ├── topdown_base_dataset.py
│ │ │ │ ├── topdown_coco_dataset.py
│ │ │ │ ├── topdown_coco_wholebody_dataset.py
│ │ │ │ ├── topdown_crowdpose_dataset.py
│ │ │ │ ├── topdown_jhmdb_dataset.py
│ │ │ │ ├── topdown_mhp_dataset.py
│ │ │ │ ├── topdown_mpii_dataset.py
│ │ │ │ ├── topdown_mpii_trb_dataset.py
│ │ │ │ ├── topdown_ochuman_dataset.py
│ │ │ │ └── topdown_posetrack18_dataset.py
│ │ ├── pipelines
│ │ │ ├── __init__.py
│ │ │ ├── bottom_up_transform.py
│ │ │ ├── hand_transform.py
│ │ │ ├── loading.py
│ │ │ ├── mesh_transform.py
│ │ │ ├── pose3d_transform.py
│ │ │ ├── shared_transform.py
│ │ │ └── top_down_transform.py
│ │ ├── registry.py
│ │ └── samplers
│ │ │ ├── __init__.py
│ │ │ └── distributed_sampler.py
│ ├── deprecated.py
│ ├── models
│ │ ├── __init__.py
│ │ ├── backbones
│ │ │ ├── __init__.py
│ │ │ ├── alexnet.py
│ │ │ ├── base_backbone.py
│ │ │ ├── cpm.py
│ │ │ ├── helpers.py
│ │ │ ├── hourglass.py
│ │ │ ├── hrnet.py
│ │ │ ├── hrt.py
│ │ │ ├── mobilenet_v2.py
│ │ │ ├── mobilenet_v3.py
│ │ │ ├── modules
│ │ │ │ ├── __init__.py
│ │ │ │ ├── basic_block.py
│ │ │ │ ├── bottleneck_block.py
│ │ │ │ ├── ffn_block.py
│ │ │ │ ├── multihead_attention.py
│ │ │ │ ├── multihead_isa_attention.py
│ │ │ │ ├── multihead_isa_pool_attention.py
│ │ │ │ └── transformer_block.py
│ │ │ ├── mspn.py
│ │ │ ├── regnet.py
│ │ │ ├── resnest.py
│ │ │ ├── resnet.py
│ │ │ ├── resnext.py
│ │ │ ├── rsn.py
│ │ │ ├── scnet.py
│ │ │ ├── seresnet.py
│ │ │ ├── seresnext.py
│ │ │ ├── shufflenet_v1.py
│ │ │ ├── shufflenet_v2.py
│ │ │ ├── tcn.py
│ │ │ ├── utils
│ │ │ │ ├── __init__.py
│ │ │ │ ├── channel_shuffle.py
│ │ │ │ ├── inverted_residual.py
│ │ │ │ ├── make_divisible.py
│ │ │ │ ├── se_layer.py
│ │ │ │ └── utils.py
│ │ │ ├── vgg.py
│ │ │ └── vit.py
│ │ ├── builder.py
│ │ ├── detectors
│ │ │ ├── __init__.py
│ │ │ ├── base.py
│ │ │ ├── bottom_up.py
│ │ │ ├── mesh.py
│ │ │ ├── multi_task.py
│ │ │ ├── pose_lifter.py
│ │ │ └── top_down.py
│ │ ├── keypoint_heads
│ │ │ ├── __init__.py
│ │ │ ├── bottom_up_higher_resolution_head.py
│ │ │ ├── bottom_up_simple_head.py
│ │ │ ├── fc_head.py
│ │ │ ├── heatmap_1d_head.py
│ │ │ ├── heatmap_3d_head.py
│ │ │ ├── multilabel_classification_head.py
│ │ │ ├── temporal_regression_head.py
│ │ │ ├── top_down_base_head.py
│ │ │ ├── top_down_multi_stage_head.py
│ │ │ └── top_down_simple_head.py
│ │ ├── losses
│ │ │ ├── __init__.py
│ │ │ ├── classfication_loss.py
│ │ │ ├── mesh_loss.py
│ │ │ ├── mse_loss.py
│ │ │ ├── multi_loss_factory.py
│ │ │ └── regression_loss.py
│ │ ├── mesh_heads
│ │ │ ├── __init__.py
│ │ │ ├── discriminator.py
│ │ │ ├── geometric_layers.py
│ │ │ └── hmr_head.py
│ │ ├── necks
│ │ │ ├── __init__.py
│ │ │ └── gap_neck.py
│ │ ├── registry.py
│ │ └── utils
│ │ │ ├── __init__.py
│ │ │ └── ops.py
│ └── utils
│ │ ├── __init__.py
│ │ ├── collect_env.py
│ │ ├── hooks.py
│ │ └── logger.py
├── requirements.txt
├── requirements
│ ├── build.txt
│ ├── docs.txt
│ ├── optional.txt
│ ├── readthedocs.txt
│ ├── runtime.txt
│ └── tests.txt
├── resources
│ └── mmpose-logo.png
├── run_dist.sh
├── setup.cfg
├── setup.py
├── tests
│ ├── data
│ │ ├── 300w
│ │ │ ├── indoor_020.png
│ │ │ ├── indoor_029.png
│ │ │ └── test_300w.json
│ │ ├── aflw
│ │ │ ├── image04476.jpg
│ │ │ ├── image22568.jpg
│ │ │ └── test_aflw.json
│ │ ├── aic
│ │ │ ├── 054d9ce9201beffc76e5ff2169d2af2f027002ca.jpg
│ │ │ ├── fa436c914fe4a8ec1ec5474af4d3820b84d17561.jpg
│ │ │ ├── ff945ae2e729f24eea992814639d59b3bdec8bd8.jpg
│ │ │ └── test_aic.json
│ │ ├── atrw
│ │ │ ├── 000061.jpg
│ │ │ ├── 003464.jpg
│ │ │ └── test_atrw.json
│ │ ├── coco
│ │ │ ├── 000000000785.jpg
│ │ │ ├── 000000040083.jpg
│ │ │ ├── 000000196141.jpg
│ │ │ ├── 000000197388.jpg
│ │ │ ├── test_coco.json
│ │ │ ├── test_coco_det_AP_H_56.json
│ │ │ └── test_coco_wholebody.json
│ │ ├── cofw
│ │ │ ├── 001766.jpg
│ │ │ ├── 001805.jpg
│ │ │ └── test_cofw.json
│ │ ├── crowdpose
│ │ │ ├── 103319.jpg
│ │ │ ├── 106848.jpg
│ │ │ ├── test_crowdpose.json
│ │ │ └── test_crowdpose_det_AP_40.json
│ │ ├── fld
│ │ │ ├── img_00000128.jpg
│ │ │ ├── img_00000132.jpg
│ │ │ └── test_fld.json
│ │ ├── fly
│ │ │ ├── 1400.jpg
│ │ │ ├── 1450.jpg
│ │ │ └── test_fly.json
│ │ ├── freihand
│ │ │ ├── 00000355.jpg
│ │ │ ├── 00017620.jpg
│ │ │ ├── 00032915.jpg
│ │ │ ├── 00050180.jpg
│ │ │ ├── 00065475.jpg
│ │ │ ├── 00082740.jpg
│ │ │ ├── 00098035.jpg
│ │ │ ├── 00115300.jpg
│ │ │ └── test_freihand.json
│ │ ├── h36m
│ │ │ ├── BF_IUV_gt
│ │ │ │ ├── S1_Directions_1.54138969_000001_467_466.png
│ │ │ │ ├── S5_SittingDown.54138969_002061_478_619.png
│ │ │ │ ├── S7_Greeting.55011271_000396_365_433.png
│ │ │ │ └── S8_WalkDog_1.55011271_000026_592_382.png
│ │ │ ├── S1_Directions_1.54138969_000001.jpg
│ │ │ ├── S5_SittingDown.54138969_002061.jpg
│ │ │ ├── S7_Greeting.55011271_000396.jpg
│ │ │ ├── S8_WalkDog_1.55011271_000026.jpg
│ │ │ ├── test_h36m.npz
│ │ │ └── test_h36m_body3d.npz
│ │ ├── horse10
│ │ │ ├── 0244.png
│ │ │ ├── 0292.png
│ │ │ ├── 0465.png
│ │ │ └── test_horse10.json
│ │ ├── interhand2.6m
│ │ │ ├── image2017.jpg
│ │ │ ├── image29590.jpg
│ │ │ ├── image44669.jpg
│ │ │ ├── image69148.jpg
│ │ │ ├── test_interhand2.6m_camera.json
│ │ │ ├── test_interhand2.6m_data.json
│ │ │ └── test_interhand2.6m_joint_3d.json
│ │ ├── jhmdb
│ │ │ ├── Frisbee_catch_f_cm_np1_ri_med_0
│ │ │ │ └── 00001.png
│ │ │ ├── Frisbee_catch_f_cm_np1_ri_med_1
│ │ │ │ └── 00001.png
│ │ │ ├── Goalkeeper_Training_Day_@_7_catch_f_cm_np1_ri_med_0
│ │ │ │ └── 00001.png
│ │ │ └── test_jhmdb_sub1.json
│ │ ├── locust
│ │ │ ├── 630.jpg
│ │ │ ├── 650.jpg
│ │ │ └── test_locust.json
│ │ ├── macaque
│ │ │ ├── PRI_1473.jpg
│ │ │ ├── d47f1b1ee9d3217e.jpg
│ │ │ └── test_macaque.json
│ │ ├── mhp
│ │ │ ├── 10084.jpg
│ │ │ ├── 10112.jpg
│ │ │ └── test_mhp.json
│ │ ├── mosh
│ │ │ └── test_mosh.npz
│ │ ├── mpii
│ │ │ ├── 004645041.jpg
│ │ │ ├── 005808361.jpg
│ │ │ ├── 051423444.jpg
│ │ │ ├── 052475643.jpg
│ │ │ ├── 060754485.jpg
│ │ │ ├── test_mpii.json
│ │ │ └── test_mpii_trb.json
│ │ ├── ochuman
│ │ │ ├── 000817.jpg
│ │ │ ├── 003799.jpg
│ │ │ ├── 003896.jpg
│ │ │ └── test_ochuman.json
│ │ ├── onehand10k
│ │ │ ├── 1402.jpg
│ │ │ ├── 33.jpg
│ │ │ ├── 784.jpg
│ │ │ ├── 9.jpg
│ │ │ └── test_onehand10k.json
│ │ ├── panoptic
│ │ │ ├── 005880453_01_l.jpg
│ │ │ ├── 005880453_01_r.jpg
│ │ │ ├── ex2_2.flv_000040_l.jpg
│ │ │ ├── ex2_2.flv_000040_r.jpg
│ │ │ └── test_panoptic.json
│ │ ├── posetrack18
│ │ │ ├── images
│ │ │ │ └── val
│ │ │ │ │ ├── 003418_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ │ │ ├── 009473_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ │ │ └── 012834_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ ├── mask
│ │ │ │ └── val
│ │ │ │ │ ├── 003418_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ │ │ ├── 009473_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ │ │ └── 012834_mpii_test
│ │ │ │ │ └── 000000.jpg
│ │ │ ├── test_posetrack18.json
│ │ │ └── test_posetrack18_human_detections.json
│ │ ├── rhd
│ │ │ ├── 00111.png
│ │ │ ├── 01111.png
│ │ │ ├── 11111.png
│ │ │ └── test_rhd.json
│ │ ├── smpl
│ │ │ └── smpl_mean_params.npz
│ │ ├── wflw
│ │ │ ├── 36_Football_americanfootball_ball_36_415.jpg
│ │ │ ├── 7_Cheering_Cheering_7_16.jpg
│ │ │ └── test_wflw.json
│ │ └── zebra
│ │ │ ├── 810.jpg
│ │ │ ├── 850.jpg
│ │ │ └── test_zebra.json
│ ├── post_processing
│ │ ├── test_group.py
│ │ └── test_nms.py
│ ├── test_api
│ │ ├── test_inference.py
│ │ └── test_inference_tracking.py
│ ├── test_backbones
│ │ ├── test_alexnet.py
│ │ ├── test_backbones_utils.py
│ │ ├── test_cpm.py
│ │ ├── test_hourglass.py
│ │ ├── test_hrnet.py
│ │ ├── test_mobilenet_v2.py
│ │ ├── test_mobilenet_v3.py
│ │ ├── test_mspn.py
│ │ ├── test_regnet.py
│ │ ├── test_resnest.py
│ │ ├── test_resnet.py
│ │ ├── test_resnext.py
│ │ ├── test_rsn.py
│ │ ├── test_scnet.py
│ │ ├── test_seresnet.py
│ │ ├── test_seresnext.py
│ │ ├── test_shufflenet_v1.py
│ │ ├── test_shufflenet_v2.py
│ │ ├── test_tcn.py
│ │ └── test_vgg.py
│ ├── test_compose.py
│ ├── test_config.py
│ ├── test_datasets
│ │ ├── test_animal_dataset.py
│ │ ├── test_body3d_dataset.py
│ │ ├── test_bottom_up_dataset.py
│ │ ├── test_face_dataset.py
│ │ ├── test_fashion_dataset.py
│ │ ├── test_hand_dataset.py
│ │ ├── test_mesh_dataset.py
│ │ └── test_top_down_dataset.py
│ ├── test_eval_hook.py
│ ├── test_evaluation
│ │ ├── test_bottom_up_eval.py
│ │ ├── test_mesh_eval.py
│ │ └── test_top_down_eval.py
│ ├── test_fp16.py
│ ├── test_loss
│ │ ├── test_bottom_up_losses.py
│ │ ├── test_classification_loss.py
│ │ ├── test_mesh_losses.py
│ │ ├── test_regression_losses.py
│ │ └── test_top_down_losses.py
│ ├── test_model
│ │ ├── test_bottom_up_forward.py
│ │ ├── test_bottom_up_head.py
│ │ ├── test_heatmap_1d_head.py
│ │ ├── test_heatmap_3d_head.py
│ │ ├── test_layer.py
│ │ ├── test_mesh_forward.py
│ │ ├── test_mesh_head.py
│ │ ├── test_multilabel_classification_head.py
│ │ ├── test_multitask_forward.py
│ │ ├── test_pose_lifter_forward.py
│ │ ├── test_temporal_regression_head.py
│ │ ├── test_top_down_forward.py
│ │ └── test_top_down_head.py
│ ├── test_necks
│ │ └── test_gap_neck.py
│ ├── test_onnx.py
│ ├── test_optimizer.py
│ ├── test_pipelines
│ │ ├── test_bottom_up_pipelines.py
│ │ ├── test_hand_transform.py
│ │ ├── test_mesh_pipelines.py
│ │ ├── test_pose3d_transform.py
│ │ ├── test_shared_transform.py
│ │ └── test_top_down_pipelines.py
│ ├── test_post_processing.py
│ ├── test_regularization.py
│ ├── test_utils.py
│ └── test_version.py
└── tools
│ ├── analysis
│ ├── analyze_logs.py
│ ├── get_flops.py
│ └── print_config.py
│ ├── benchmark_inference.py
│ ├── benchmark_processing.py
│ ├── dataset
│ ├── parse_cofw_dataset.py
│ ├── parse_deepposekit_dataset.py
│ ├── parse_macaquepose_dataset.py
│ └── preprocess_h36m.py
│ ├── dist_test.sh
│ ├── dist_train.sh
│ ├── mat2json.py
│ ├── publish_model.py
│ ├── pytorch2onnx.py
│ ├── slurm_test.sh
│ ├── slurm_train.sh
│ ├── test.py
│ └── train.py
└── seg
├── .gitignore
├── GETTING_STARTED.md
├── LICENSE
├── MODEL_ZOO.md
├── README.md
├── README_OPENSEG.md
├── config.profile
├── configs
├── ade20k
│ ├── H_48_D_4.json
│ ├── H_48_D_4_RMI.json
│ ├── H_48_D_4_TEST.json
│ ├── H_48_D_4_TEST_MS3x.json
│ ├── H_SEGFIX.json
│ ├── M_V2_D_8.json
│ ├── R_101_D_8.json
│ ├── R_101_D_8_TEST.json
│ ├── R_50_D_8.json
│ └── W_38_D_8.json
├── celeba
│ ├── H_48_D_4.json
│ └── H_48_D_4_TEST.json
├── cityscapes
│ ├── H_48_D_4.json
│ ├── H_48_D_4_RMI.json
│ ├── H_48_D_4_RMI_V2.json
│ ├── H_48_D_4_SSV1.json
│ ├── H_48_D_4_SSV2.json
│ ├── H_48_D_4_TEST.json
│ ├── H_48_D_4_TEST_DEPTH.json
│ ├── H_48_D_4_TEST_MS3x.json
│ ├── H_48_D_4_TEST_MS6x.json
│ ├── H_48_D_4_WD1E4.json
│ ├── H_64_D_4.json
│ ├── H_SEGFIX.json
│ ├── M_V2_D_8.json
│ ├── R_101_D_16.json
│ ├── R_101_D_8.json
│ ├── R_101_D_8_EDGE_VOID.json
│ ├── R_101_D_8_NON_EDGE_VOID.json
│ ├── R_101_D_8_TEST.json
│ ├── R_18_D_8.json
│ ├── R_50_D_8.json
│ ├── W_38_D_8.json
│ ├── X_65_D_16.json
│ └── X_65_D_8.json
├── coco_stuff
│ ├── H_48_D_4.json
│ ├── H_48_D_4_RMI.json
│ ├── H_48_D_4_TEST.json
│ ├── H_48_D_4_TEST_MS3x.json
│ ├── R_101_D_8.json
│ └── R_101_D_8_TEST.json
├── lip
│ ├── H_48_D_4.json
│ ├── H_48_D_4_RMI.json
│ ├── H_48_D_4_TEST.json
│ ├── R_101_D_16.json
│ ├── R_101_D_8.json
│ └── W_38_D_8.json
├── mapillary
│ └── H_48_D_4_1024x1024.json
├── pascal_context
│ ├── H_48_D_4.json
│ ├── H_48_D_4_RMI.json
│ ├── H_48_D_4_TEST.json
│ ├── H_48_D_4_TEST_MS3x.json
│ ├── R_101_D_8.json
│ ├── R_101_D_8_TEST.json
│ └── W_38_D_8.json
└── segfix
│ └── H_SEGFIX.json
├── imgs
├── HRT_arch5.png
├── OCR.PNG
└── SegFix.PNG
├── lib
├── __init__.py
├── datasets
│ ├── __init__.py
│ ├── data_loader.py
│ ├── loader
│ │ ├── __init__.py
│ │ ├── ade20k_loader.py
│ │ ├── default_loader.py
│ │ ├── lip_loader.py
│ │ ├── multi_dataset_loader.py
│ │ └── offset_loader.py
│ ├── preprocess
│ │ ├── README.md
│ │ ├── __init__.py
│ │ ├── ade20k
│ │ │ ├── __init__.py
│ │ │ ├── ade20k_generator.py
│ │ │ ├── ade20k_generator.sh
│ │ │ └── dt_offset_generator.py
│ │ ├── cityscapes
│ │ │ ├── __init__.py
│ │ │ ├── cityscapes_generator.py
│ │ │ ├── cityscapes_instance_generator.py
│ │ │ ├── dt_offset_generator.py
│ │ │ ├── edge_generator.py
│ │ │ ├── instance_dt_offset_generator.py
│ │ │ └── instance_edge_generator.py
│ │ ├── coco_stuff
│ │ │ └── coco_stuff_generator.py
│ │ ├── face
│ │ │ ├── celebmask_color.py
│ │ │ ├── celebmask_label_generator.py
│ │ │ ├── celebmask_partition.py
│ │ │ ├── celebmask_resize.py
│ │ │ └── prepare_celeb.sh
│ │ ├── lip
│ │ │ ├── __init__.py
│ │ │ └── lip.py
│ │ ├── mapillary
│ │ │ ├── mapillary_generator.py
│ │ │ └── mapillary_generator.sh
│ │ ├── pascal_context
│ │ │ ├── pascal_context_generator.py
│ │ │ └── pascal_context_generator.sh
│ │ └── pascal_voc
│ │ │ └── pascal_voc_generator.py
│ └── tools
│ │ ├── __init__.py
│ │ ├── collate.py
│ │ ├── cv2_aug_transforms.py
│ │ ├── pil_aug_transforms.py
│ │ └── transforms.py
├── extensions
│ ├── __init__.py
│ ├── cc_attention
│ │ ├── __init__.py
│ │ ├── _ext
│ │ │ └── __init__.py
│ │ ├── build.py
│ │ ├── build.sh
│ │ ├── functions.py
│ │ └── src
│ │ │ ├── ca.cu
│ │ │ ├── ca.h
│ │ │ ├── common.h
│ │ │ ├── lib_cffi.cpp
│ │ │ └── lib_cffi.h
│ ├── crf
│ │ └── dense_crf.py
│ ├── dcn
│ │ ├── README.md
│ │ ├── __init__.py
│ │ ├── _ext
│ │ │ ├── __init__.py
│ │ │ ├── deform_conv
│ │ │ │ └── __init__.py
│ │ │ └── modulated_dcn
│ │ │ │ └── __init__.py
│ │ ├── build.py
│ │ ├── build_modulated.py
│ │ ├── functions
│ │ │ ├── __init__.py
│ │ │ ├── deform_conv.py
│ │ │ └── modulated_dcn_func.py
│ │ ├── make.sh
│ │ ├── make_p100.sh
│ │ ├── modules
│ │ │ ├── __init__.py
│ │ │ ├── deform_conv.py
│ │ │ └── modulated_dcn.py
│ │ ├── src
│ │ │ ├── cuda
│ │ │ │ ├── deform_psroi_pooling_cuda.cu
│ │ │ │ ├── deform_psroi_pooling_cuda.h
│ │ │ │ ├── modulated_deform_im2col_cuda.cu
│ │ │ │ └── modulated_deform_im2col_cuda.h
│ │ │ ├── deform_conv.c
│ │ │ ├── deform_conv.h
│ │ │ ├── deform_conv_cuda.c
│ │ │ ├── deform_conv_cuda.h
│ │ │ ├── deform_conv_cuda_kernel.cu
│ │ │ ├── deform_conv_cuda_kernel.h
│ │ │ ├── modulated_dcn.c
│ │ │ ├── modulated_dcn.h
│ │ │ ├── modulated_dcn_cuda.c
│ │ │ └── modulated_dcn_cuda.h
│ │ ├── test.py
│ │ └── test_modulated.py
│ ├── frn
│ │ ├── __init__.py
│ │ └── frn.py
│ ├── inplace_abn
│ │ ├── __init__.py
│ │ ├── bn.py
│ │ ├── functions.py
│ │ └── src
│ │ │ ├── common.h
│ │ │ ├── inplace_abn.cpp
│ │ │ ├── inplace_abn.h
│ │ │ ├── inplace_abn_cpu.cpp
│ │ │ └── inplace_abn_cuda.cu
│ ├── inplace_abn_1
│ │ ├── __init__.py
│ │ ├── bn.py
│ │ ├── functions.py
│ │ ├── misc.py
│ │ └── src
│ │ │ ├── checks.h
│ │ │ ├── inplace_abn.cpp
│ │ │ ├── inplace_abn.h
│ │ │ ├── inplace_abn_cpu.cpp
│ │ │ ├── inplace_abn_cuda.cu
│ │ │ ├── inplace_abn_cuda_half.cu
│ │ │ └── utils
│ │ │ ├── checks.h
│ │ │ ├── common.h
│ │ │ └── cuda.cuh
│ ├── pacnet
│ │ ├── __init__.py
│ │ ├── pac.py
│ │ ├── paccrf.py
│ │ └── test_pac.py
│ ├── parallel
│ │ ├── __init__.py
│ │ ├── _functions.py
│ │ ├── data_container.py
│ │ ├── data_parallel.py
│ │ ├── distributed.py
│ │ └── scatter_gather.py
│ ├── switchablenorms
│ │ ├── __init__.py
│ │ └── switchable_norm.py
│ └── syncbn
│ │ ├── __init__.py
│ │ ├── allreduce.py
│ │ ├── comm.py
│ │ ├── module.py
│ │ └── src
│ │ ├── common.h
│ │ ├── device_tensor.h
│ │ ├── operator.cpp
│ │ ├── operator.h
│ │ ├── syncbn_cpu.cpp
│ │ └── syncbn_kernel.cu
├── loss
│ ├── __init__.py
│ ├── loss_helper.py
│ ├── loss_manager.py
│ └── rmi_loss.py
├── metrics
│ ├── F1_running_score.py
│ ├── __init__.py
│ ├── ade20k_evaluator.py
│ ├── cityscapes
│ │ ├── __init__.py
│ │ ├── evaluation
│ │ │ ├── __init__.py
│ │ │ ├── addToConfusionMatrix.c
│ │ │ ├── addToConfusionMatrix.pyx
│ │ │ ├── addToConfusionMatrix_impl.c
│ │ │ ├── csHelpers.py
│ │ │ ├── evalInstanceLevelSemanticLabeling.py
│ │ │ ├── evalPixelLevelSemanticLabeling.py
│ │ │ ├── instance.py
│ │ │ └── instances2dict.py
│ │ ├── helpers
│ │ │ ├── __init__.py
│ │ │ ├── annotation.py
│ │ │ ├── csHelpers.py
│ │ │ ├── labels.py
│ │ │ └── labels_cityPersons.py
│ │ ├── make.sh
│ │ └── setup.py
│ ├── cityscapes_evaluator.py
│ ├── cocostuff_evaluator.py
│ ├── pascal_context_evaluator.py
│ ├── running_score.py
│ └── running_score_mp.py
├── models
│ ├── __init__.py
│ ├── backbones
│ │ ├── __init__.py
│ │ ├── backbone_selector.py
│ │ ├── hrnet
│ │ │ ├── __init__.py
│ │ │ ├── hrnet_backbone.py
│ │ │ └── hrnet_config.py
│ │ ├── hrt
│ │ │ ├── __init__.py
│ │ │ ├── hrt_backbone.py
│ │ │ ├── hrt_config.py
│ │ │ └── modules
│ │ │ │ ├── __init__.py
│ │ │ │ ├── bottleneck_block.py
│ │ │ │ ├── ffn_block.py
│ │ │ │ ├── multihead_attention.py
│ │ │ │ ├── multihead_isa_attention.py
│ │ │ │ ├── multihead_isa_pool_attention.py
│ │ │ │ └── transformer_block.py
│ │ ├── resnet
│ │ │ ├── __init__.py
│ │ │ ├── dcn_resnet_models.py
│ │ │ ├── resnest_models.py
│ │ │ ├── resnet_backbone.py
│ │ │ ├── resnet_models.py
│ │ │ ├── resnext_models.py
│ │ │ ├── wide_resnet_models.py
│ │ │ └── wsl_resnext_models.py
│ │ ├── swin
│ │ │ ├── swin_backbone.py
│ │ │ └── swin_config.py
│ │ └── vit
│ │ │ ├── helper.py
│ │ │ ├── position_encoding.py
│ │ │ ├── setr.py
│ │ │ └── transformer.py
│ ├── model_manager.py
│ ├── modules
│ │ ├── __init__.py
│ │ ├── asp_oc_block.py
│ │ ├── aspp_block.py
│ │ ├── base_oc_block.py
│ │ ├── decoder_block.py
│ │ ├── edge_block.py
│ │ ├── infer_time.sh
│ │ ├── isa_block.py
│ │ ├── offset_block.py
│ │ ├── psp_block.py
│ │ ├── spatial_ocr_block.py
│ │ ├── time_log
│ │ └── upernet_block.py
│ ├── nets
│ │ ├── __init__.py
│ │ ├── ce2pnet.py
│ │ ├── fcnet.py
│ │ ├── hrnet.py
│ │ ├── hrt.py
│ │ ├── ideal_ocrnet.py
│ │ ├── isanet.py
│ │ ├── ocnet.py
│ │ ├── ocrnet.py
│ │ ├── segfix.py
│ │ └── swin.py
│ └── tools
│ │ ├── __init__.py
│ │ └── module_helper.py
├── utils
│ ├── __init__.py
│ ├── distributed.py
│ ├── flops
│ │ ├── flop_count.py
│ │ └── jit_handles.py
│ ├── helpers
│ │ ├── __init__.py
│ │ ├── dc_helper.py
│ │ ├── file_helper.py
│ │ ├── image_helper.py
│ │ ├── json_helper.py
│ │ ├── mask_helper.py
│ │ ├── offset_helper.py
│ │ └── video_helper.py
│ └── tools
│ │ ├── __init__.py
│ │ ├── average_meter.py
│ │ ├── configer.py
│ │ ├── logger.py
│ │ └── timer.py
└── vis
│ ├── __init__.py
│ ├── attention_visualizer.py
│ ├── color150.mat
│ ├── color60.mat
│ ├── log_visualizer.py
│ ├── palette.py
│ ├── seg_parser.py
│ ├── seg_visualizer.py
│ └── tensor_visualizer.py
├── main.py
├── requirements.txt
├── run_dist.sh
├── run_local.sh
├── scripts
├── ade20k
│ ├── hrnet
│ │ ├── run_h_48_d_4_asp_ocr_ohem.sh
│ │ ├── run_h_48_d_4_isa_ohem.sh
│ │ ├── run_h_48_d_4_ocr_ohem.sh
│ │ ├── run_h_48_d_4_ocr_ohem_paddle.sh
│ │ ├── run_h_48_d_4_ocr_rmi_paddle.sh
│ │ └── run_h_48_d_4_train.sh
│ ├── hrt
│ │ ├── run_hrt_base_ocr_v2_ohem.sh
│ │ ├── run_hrt_base_ocr_v2_ohem_w13.sh
│ │ ├── run_hrt_base_ocr_v3_ohem_w13.sh
│ │ └── run_hrt_small_ocr_v2_ohem.sh
│ ├── isa
│ │ └── run_wideb5_isanet_ade20k.sh
│ ├── ocnet
│ │ ├── run_res101d8_aspocnet_ade20k_seg.sh
│ │ ├── run_res101d8_aspp_baseocnet_ade20k_seg.sh
│ │ ├── run_res101d8_baseocnet_ade20k_seg.sh
│ │ ├── run_res50d8_aspocnet_ade20k_seg.sh
│ │ ├── run_res50d8_aspp_baseocnet_ade20k_seg.sh
│ │ └── run_res50d8_baseocnet_ade20k_seg.sh
│ ├── ocrnet
│ │ ├── run_res101d8_aspocr_ade20k.sh
│ │ ├── run_res101d8_fastaspocnet_ade20k_seg_test.sh
│ │ ├── run_res101d8_fastbaseocnet_ade20k_seg.sh
│ │ ├── run_res101d8_fastbaseocnet_ade20k_seg_ohem.sh
│ │ ├── run_res101d8_fastbaseocnet_ade20k_trainval.sh
│ │ ├── run_res101d8_fastbaseocnet_ade20k_trainval_ohem.sh
│ │ ├── run_res101d8_ideal_ocr_ade20k.sh
│ │ ├── run_res101d8_ideal_ocr_b_ade20k.sh
│ │ ├── run_res101d8_ideal_ocr_c_ade20k.sh
│ │ ├── run_res101d8_ocr_ade20k.sh
│ │ ├── run_res50d8_fastaspocnet_ade20k_seg.sh
│ │ ├── run_res50d8_ideal_ocr_ade20k.sh
│ │ └── run_res50d8_ocr_ade20k.sh
│ ├── run_res101d8_fcn_ade20k_seg.sh
│ └── wide_hrnet
│ │ ├── run_h_32_d_4_ocr_v2_ohem_adamw_lr1e4.sh
│ │ └── run_h_78_d_4_ocr_v2_ohem_adamw_lr1e4.sh
├── celeba
│ ├── aml_run_h_48_d_4_ocr_train.sh
│ ├── aml_run_h_48_d_4_ocr_train_200k.sh
│ ├── aml_run_h_48_d_4_ocr_train_lr1e2.sh
│ ├── aml_run_h_48_d_4_ocr_train_lr1e3.sh
│ ├── aml_run_h_48_d_4_ocr_train_lr2e2.sh
│ ├── aml_run_h_48_d_4_ocr_train_lr5e3.sh
│ ├── run_h_48_d_4_ocr_train.sh
│ └── run_h_48_d_4_train.sh
├── cityscapes
│ ├── fcn
│ │ ├── run_r_101_d_8_fcn_train.sh
│ │ └── run_r_101_d_8_fcn_wo_dsn_train.sh
│ ├── hrnet
│ │ ├── run_h_48_d_4.sh
│ │ ├── run_h_48_d_4_ocr.sh
│ │ ├── run_h_48_d_4_ocr_b.sh
│ │ ├── run_h_48_d_4_ocr_b_mapillary_trainval_coarse_ohem.sh
│ │ ├── run_h_48_d_4_ocr_b_mapillary_trainval_coarse_trainval_ohem.sh
│ │ ├── run_h_48_d_4_ocr_b_mapillary_trainval_ohem.sh
│ │ ├── run_h_48_d_4_ocr_paddle.sh
│ │ ├── run_h_48_d_4_ocr_paddle_ohem.sh
│ │ ├── run_h_48_d_4_ocr_rmi_paddle.sh
│ │ ├── run_h_48_d_4_ocr_trainval.sh
│ │ ├── run_h_48_d_4_ocr_trainval_ohem_paddle.sh
│ │ └── run_h_48_d_4_ocr_trainval_paddle.sh
│ ├── hrt
│ │ ├── run_hrt_base_ocr_v2_ohem.sh
│ │ ├── run_hrt_base_ocr_v2_ohem_w15.sh
│ │ ├── run_hrt_base_ocr_v3_ohem_w15.sh
│ │ └── run_hrt_small_ocr_v2_ohem.sh
│ ├── isa
│ │ └── run_r_101_d_8_isa_train.sh
│ ├── ocnet
│ │ ├── run_r_101_d_8_aspoc_train.sh
│ │ ├── run_r_101_d_8_baseoc_train.sh
│ │ └── run_r_101_d_8_pyramidoc_train.sh
│ ├── ocrnet
│ │ ├── run_ideal_distribute_ocrnet.sh
│ │ ├── run_ideal_gather_ocrnet.sh
│ │ ├── run_ideal_spatial_ocrnet.sh
│ │ ├── run_ideal_spatial_ocrnet_b.sh
│ │ ├── run_ideal_spatial_ocrnet_c.sh
│ │ ├── run_r_101_d_8_asp_ocrnet_train.sh
│ │ ├── run_r_101_d_8_ocrnet_train.sh
│ │ ├── run_r_101_d_8_ocrnet_trainval.sh
│ │ ├── run_spatial_ocrnet_trainval_coarse.sh
│ │ ├── run_spatial_ocrnet_trainval_coarse_trainval.sh
│ │ ├── run_spatial_ocrnet_trainval_mapillary.sh
│ │ └── run_spatial_ocrnet_trainval_mapillary_coarse.sh
│ ├── resnest
│ │ └── run_r_101_d_8_fcn.sh
│ ├── segfix.py
│ ├── segfix
│ │ ├── run_h_48_d_4_segfix.sh
│ │ ├── run_h_48_d_4_segfix_inst.sh
│ │ ├── run_h_48_d_4_segfix_trainval.sh
│ │ ├── run_hx_20_d_2_segfix.sh
│ │ ├── run_hx_20_d_2_segfix_inst.sh
│ │ └── run_hx_20_d_2_segfix_trainval.sh
│ ├── segfix_ade20k.py
│ ├── segfix_instance.py
│ └── swin
│ │ └── run_swin_tiny_swin_uper_ohem_adamw_lr1e4.sh
├── coco_stuff
│ ├── hrt
│ │ ├── run_hrt_base_ocr_v2_ohem.sh
│ │ ├── run_hrt_base_ocr_v2_ohem_w15.sh
│ │ ├── run_hrt_base_ocr_v3_ohem_w15.sh
│ │ └── run_hrt_small_ocr_v2_ohem.sh
│ ├── run_h_48_d_4_isa_train.sh
│ ├── run_h_48_d_4_ocr_ohem_train.sh
│ ├── run_h_48_d_4_ocr_ohem_train_paddle.sh
│ ├── run_h_48_d_4_ocr_ohem_train_rmi_paddle.sh
│ ├── run_h_48_d_4_ocr_train.sh
│ ├── run_h_48_d_4_train.sh
│ ├── run_r_101_d_8_gt_ocr_train.sh
│ ├── run_r_101_d_8_ocr_train.sh
│ └── run_r_101_d_8_train.sh
├── lip
│ ├── run_h_48_d_4_isa_train.sh
│ ├── run_h_48_d_4_ocr_b_train.sh
│ ├── run_h_48_d_4_ocr_train.sh
│ ├── run_h_48_d_4_ocr_train_paddle.sh
│ ├── run_h_48_d_4_ocr_train_rmi_paddle.sh
│ ├── run_h_48_d_4_train.sh
│ ├── run_r_101_d_16_ce2p_gt_ocrnet_train.sh
│ └── run_r_101_d_16_ce2p_ocr_train.sh
├── mapillary
│ └── run_h_48_d_4_ocr_b.sh
├── pascal_context
│ ├── hrt
│ │ ├── run_hrt_base_ocr_v2_ohem.sh
│ │ ├── run_hrt_base_ocr_v2_ohem_w15.sh
│ │ ├── run_hrt_base_ocr_v3_ohem_w15.sh
│ │ └── run_hrt_small_ocr_v2_ohem.sh
│ ├── run_h_48_d_4_isa_train.sh
│ ├── run_h_48_d_4_ocr_b_train.sh
│ ├── run_h_48_d_4_ocr_train.sh
│ ├── run_h_48_d_4_ocr_train_paddle.sh
│ ├── run_h_48_d_4_ocr_train_rmi_paddle.sh
│ ├── run_h_48_d_4_segtr_train.sh
│ ├── run_h_48_d_4_segtre_train.sh
│ ├── run_h_48_d_4_train.sh
│ ├── run_r_101_d_8_aspocr_train.sh
│ ├── run_r_101_d_8_baseoc_train.sh
│ ├── run_r_101_d_8_gt_ocr_train.sh
│ ├── run_r_101_d_8_ocr_train.sh
│ └── run_r_101_d_8_train.sh
├── segfix
│ └── run_hx_20_d_2_cityscapes_ade20k.sh
└── test_flops
│ ├── test_hrnet_w48_ocr_1024x1024.sh
│ ├── test_hrnet_w48_ocr_512x512.sh
│ ├── test_hrvit_w18_ocr_v2_1024x1024.sh
│ ├── test_hrvit_w32_ocr_v2_1024x1024.sh
│ ├── test_hrvit_w32_ocr_v2_512x512.sh
│ ├── test_hrvit_w78_ocr_v2_1024x1024.sh
│ ├── test_hrvit_w78_ocr_v2_512x512.sh
│ ├── test_wide_hrnet_w32_ocr_v2_1024x1024.sh
│ ├── test_wide_hrnet_w32_ocr_v2_512x512.sh
│ ├── test_wide_hrnet_w78_ocr_v2_1024x1024.sh
│ └── test_wide_hrnet_w78_ocr_v2_512x512.sh
└── segmentor
├── __init__.py
├── tester.py
├── tester_flops.py
├── tester_offset.py
├── tools
├── __init__.py
├── blob_helper.py
├── cost_helper.py
├── data_helper.py
├── evaluator
│ ├── __init__.py
│ ├── base.py
│ ├── standard.py
│ └── tasks.py
├── module_runner.py
└── optim_scheduler.py
└── trainer.py
/HRFormer-20-fps.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/HRFormer-20-fps.gif
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | MIT License
2 |
3 | Copyright (c) 2021 HRT-group (Yuhui Yuan,Rao Fu,Lang Huang)
4 |
5 | Permission is hereby granted, free of charge, to any person obtaining a copy
6 | of this software and associated documentation files (the "Software"), to deal
7 | in the Software without restriction, including without limitation the rights
8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 |
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 |
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.
22 |
--------------------------------------------------------------------------------
/cls/.gitattributes:
--------------------------------------------------------------------------------
1 | ckpt_epoch_11.pth filter=lfs diff=lfs merge=lfs -text
2 |
--------------------------------------------------------------------------------
/cls/LICENSE:
--------------------------------------------------------------------------------
1 | MIT License
2 |
3 | Copyright (c) Microsoft Corporation.
4 |
5 | Permission is hereby granted, free of charge, to any person obtaining a copy
6 | of this software and associated documentation files (the "Software"), to deal
7 | in the Software without restriction, including without limitation the rights
8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 |
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 |
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE
22 |
--------------------------------------------------------------------------------
/cls/configs/deit/deit_base_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: deit
3 | NAME: deit_base_patch16_224
4 | DEIT:
5 | PATCH_SIZE: 16
6 | EMBED_DIM: 768
7 | DEPTHS: 12
8 | NUM_HEADS: 12
9 | MLP_RATIO: 4
10 | QKV_BIAS: True
11 | DROP_PATH_RATE: 0.5
12 |
--------------------------------------------------------------------------------
/cls/configs/deit/deit_small_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: deit
3 | NAME: deit_small_patch16_224
4 | DEIT:
5 | PATCH_SIZE: 16
6 | EMBED_DIM: 384
7 | DEPTHS: 12
8 | NUM_HEADS: 6
9 | MLP_RATIO: 4
10 | QKV_BIAS: True
11 | DROP_PATH_RATE: 0.3
--------------------------------------------------------------------------------
/cls/configs/deit/deit_tiny_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: deit
3 | NAME: deit_tiny_patch16_224
4 | DEIT:
5 | DROP_PATH_RATE: 0.1
6 | PATCH_SIZE: 16
7 | EMBED_DIM: 192
8 | DEPTHS: 12
9 | NUM_HEADS: 3
10 | MLP_RATIO: 4
11 | QKV_BIAS: True
--------------------------------------------------------------------------------
/cls/configs/hrnet/hrnet_w18.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: hrnet
3 | NAME: hrnet_w18
4 | NUM_CLASSES: 1000
5 | HRNET:
6 | STAGE1:
7 | NUM_MODULES: 1
8 | NUM_BRANCHES: 1
9 | NUM_BLOCKS: [4]
10 | NUM_CHANNELS: [64]
11 | BLOCK: 'BOTTLENECK'
12 |
13 | STAGE2:
14 | NUM_MODULES: 1
15 | NUM_BRANCHES: 2
16 | NUM_BLOCKS: [4, 4]
17 | NUM_CHANNELS: [18, 36]
18 | BLOCK: 'BASIC'
19 |
20 | STAGE3:
21 | NUM_MODULES: 4
22 | NUM_BRANCHES: 3
23 | NUM_BLOCKS: [4, 4, 4]
24 | NUM_CHANNELS: [18, 36, 72]
25 | BLOCK: 'BASIC'
26 |
27 | STAGE4:
28 | NUM_MODULES: 3
29 | NUM_BRANCHES: 4
30 | NUM_BLOCKS: [4, 4, 4, 4]
31 | NUM_CHANNELS: [18, 36, 72, 144]
32 | BLOCK: 'BASIC'
--------------------------------------------------------------------------------
/cls/configs/hrnet/hrnet_w18_small_v2.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: hrnet
3 | NAME: hrnet_w18_small_v2
4 | NUM_CLASSES: 1000
5 | HRNET:
6 | STAGE1:
7 | NUM_MODULES: 1
8 | NUM_BRANCHES: 1
9 | NUM_BLOCKS: [2]
10 | NUM_CHANNELS: [64]
11 | BLOCK: 'BOTTLENECK'
12 |
13 | STAGE2:
14 | NUM_MODULES: 1
15 | NUM_BRANCHES: 2
16 | NUM_BLOCKS: [2, 2]
17 | NUM_CHANNELS: [18, 36]
18 | BLOCK: 'BASIC'
19 |
20 | STAGE3:
21 | NUM_MODULES: 3
22 | NUM_BRANCHES: 3
23 | NUM_BLOCKS: [2, 2, 2]
24 | NUM_CHANNELS: [18, 36, 72]
25 | BLOCK: 'BASIC'
26 |
27 | STAGE4:
28 | NUM_MODULES: 2
29 | NUM_BRANCHES: 4
30 | NUM_BLOCKS: [2, 2, 2, 2]
31 | NUM_CHANNELS: [18, 36, 72, 144]
32 | BLOCK: 'BASIC'
--------------------------------------------------------------------------------
/cls/configs/hrnet/hrnet_w32.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: hrnet
3 | NAME: hrnet_w32
4 | NUM_CLASSES: 1000
5 | HRNET:
6 | STAGE1:
7 | NUM_MODULES: 1
8 | NUM_BRANCHES: 1
9 | NUM_BLOCKS: [4]
10 | NUM_CHANNELS: [64]
11 | BLOCK: 'BOTTLENECK'
12 |
13 | STAGE2:
14 | NUM_MODULES: 1
15 | NUM_BRANCHES: 2
16 | NUM_BLOCKS: [4, 4]
17 | NUM_CHANNELS: [32, 64]
18 | BLOCK: 'BASIC'
19 |
20 | STAGE3:
21 | NUM_MODULES: 4
22 | NUM_BRANCHES: 3
23 | NUM_BLOCKS: [4, 4, 4]
24 | NUM_CHANNELS: [32, 64, 128]
25 | BLOCK: 'BASIC'
26 |
27 | STAGE4:
28 | NUM_MODULES: 3
29 | NUM_BRANCHES: 4
30 | NUM_BLOCKS: [4, 4, 4, 4]
31 | NUM_CHANNELS: [32, 64, 128, 256]
32 | BLOCK: 'BASIC'
--------------------------------------------------------------------------------
/cls/configs/hrnet/hrnet_w48.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: hrnet
3 | NAME: hrnet_w48
4 | NUM_CLASSES: 1000
5 | HRNET:
6 | STAGE1:
7 | NUM_MODULES: 1
8 | NUM_BRANCHES: 1
9 | NUM_BLOCKS: [4]
10 | NUM_CHANNELS: [64]
11 | BLOCK: 'BOTTLENECK'
12 |
13 | STAGE2:
14 | NUM_MODULES: 1
15 | NUM_BRANCHES: 2
16 | NUM_BLOCKS: [4, 4]
17 | NUM_CHANNELS: [48, 96]
18 | BLOCK: 'BASIC'
19 |
20 | STAGE3:
21 | NUM_MODULES: 4
22 | NUM_BRANCHES: 3
23 | NUM_BLOCKS: [4, 4, 4]
24 | NUM_CHANNELS: [48, 96, 192]
25 | BLOCK: 'BASIC'
26 |
27 | STAGE4:
28 | NUM_MODULES: 3
29 | NUM_BRANCHES: 4
30 | NUM_BLOCKS: [4, 4, 4, 4]
31 | NUM_CHANNELS: [48, 96, 192, 384]
32 | BLOCK: 'BASIC'
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b0_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b0_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.1
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [32, 64, 160, 256]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [8, 8, 4, 4]
10 | DEPTHS: [2, 2, 2, 2]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b1_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b1_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.1
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [64, 128, 320, 512]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [8, 8, 4, 4]
10 | DEPTHS: [2, 2, 2, 2]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b2_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b2_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.2
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [64, 128, 320, 512]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [8, 8, 4, 4]
10 | DEPTHS: [3, 4, 6, 3]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b3_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b3_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.3
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [64, 128, 320, 512]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [8, 8, 4, 4]
10 | DEPTHS: [3, 4, 18, 3]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b4_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b4_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.4
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [64, 128, 320, 512]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [8, 8, 4, 4]
10 | DEPTHS: [3, 8, 27, 3]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/pvt/pvt_v2_b5_patch16_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: pvt_v2
3 | NAME: pvt_v2_b5_patch16_224
4 | PVT:
5 | DROP_PATH_RATE: 0.5
6 | PATCH_SIZE: 16
7 | EMBED_DIMS: [64, 128, 320, 512]
8 | NUM_HEADS: [1, 2, 5, 8]
9 | MLP_RATIOS: [4, 4, 4, 4]
10 | DEPTHS: [3, 6, 40, 3]
11 | SR_RATIOS: [8, 4, 2, 1]
--------------------------------------------------------------------------------
/cls/configs/swin/swin_base_patch4_window12_384.yaml:
--------------------------------------------------------------------------------
1 | # only for evaluation
2 | DATA:
3 | IMG_SIZE: 384
4 | MODEL:
5 | TYPE: swin
6 | NAME: swin_base_patch4_window12_384
7 | SWIN:
8 | EMBED_DIM: 128
9 | DEPTHS: [ 2, 2, 18, 2 ]
10 | NUM_HEADS: [ 4, 8, 16, 32 ]
11 | WINDOW_SIZE: 12
12 | TEST:
13 | CROP: False
--------------------------------------------------------------------------------
/cls/configs/swin/swin_base_patch4_window7_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: swin
3 | NAME: swin_base_patch4_window7_224
4 | DROP_PATH_RATE: 0.5
5 | SWIN:
6 | EMBED_DIM: 128
7 | DEPTHS: [ 2, 2, 18, 2 ]
8 | NUM_HEADS: [ 4, 8, 16, 32 ]
9 | WINDOW_SIZE: 7
--------------------------------------------------------------------------------
/cls/configs/swin/swin_large_patch4_window12_384.yaml:
--------------------------------------------------------------------------------
1 | # only for evaluation
2 | DATA:
3 | IMG_SIZE: 384
4 | MODEL:
5 | TYPE: swin
6 | NAME: swin_large_patch4_window12_384
7 | SWIN:
8 | EMBED_DIM: 192
9 | DEPTHS: [ 2, 2, 18, 2 ]
10 | NUM_HEADS: [ 6, 12, 24, 48 ]
11 | WINDOW_SIZE: 12
12 | TEST:
13 | CROP: False
--------------------------------------------------------------------------------
/cls/configs/swin/swin_large_patch4_window7_224.yaml:
--------------------------------------------------------------------------------
1 | # only for evaluation
2 | MODEL:
3 | TYPE: swin
4 | NAME: swin_large_patch4_window7_224
5 | SWIN:
6 | EMBED_DIM: 192
7 | DEPTHS: [ 2, 2, 18, 2 ]
8 | NUM_HEADS: [ 6, 12, 24, 48 ]
9 | WINDOW_SIZE: 7
--------------------------------------------------------------------------------
/cls/configs/swin/swin_small_patch4_window7_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: swin
3 | NAME: swin_small_patch4_window7_224
4 | DROP_PATH_RATE: 0.3
5 | SWIN:
6 | EMBED_DIM: 96
7 | DEPTHS: [ 2, 2, 18, 2 ]
8 | NUM_HEADS: [ 3, 6, 12, 24 ]
9 | WINDOW_SIZE: 7
--------------------------------------------------------------------------------
/cls/configs/swin/swin_tiny_patch4_window7_224.yaml:
--------------------------------------------------------------------------------
1 | MODEL:
2 | TYPE: swin
3 | NAME: swin_tiny_patch4_window7_224
4 | DROP_PATH_RATE: 0.2
5 | NUM_CLASSES: 1000
6 | SWIN:
7 | EMBED_DIM: 96
8 | DEPTHS: [ 2, 2, 6, 2 ]
9 | NUM_HEADS: [ 3, 6, 12, 24 ]
10 | WINDOW_SIZE: 7
--------------------------------------------------------------------------------
/cls/data/__init__.py:
--------------------------------------------------------------------------------
1 | from .build import build_loader
--------------------------------------------------------------------------------
/cls/data/samplers.py:
--------------------------------------------------------------------------------
1 | # --------------------------------------------------------
2 | # High Resolution Transformer
3 | # Copyright (c) 2021 Microsoft
4 | # Licensed under The MIT License [see LICENSE for details]
5 | # Written by Ze Liu
6 | # Modified by Rao Fu, RainbowSecret
7 | # --------------------------------------------------------
8 |
9 | import torch
10 |
11 |
12 | class SubsetRandomSampler(torch.utils.data.Sampler):
13 | r"""Samples elements randomly from a given list of indices, without replacement.
14 |
15 | Arguments:
16 | indices (sequence): a sequence of indices
17 | """
18 |
19 | def __init__(self, indices):
20 | self.epoch = 0
21 | self.indices = indices
22 |
23 | def __iter__(self):
24 | return (self.indices[i] for i in torch.randperm(len(self.indices)))
25 |
26 | def __len__(self):
27 | return len(self.indices)
28 |
29 | def set_epoch(self, epoch):
30 | self.epoch = epoch
31 |
--------------------------------------------------------------------------------
/cls/figures/HRFormer.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRFormer.png
--------------------------------------------------------------------------------
/cls/figures/HRFormerUnit.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRFormerUnit.png
--------------------------------------------------------------------------------
/cls/figures/HRT_arch5.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRT_arch5.pdf
--------------------------------------------------------------------------------
/cls/figures/HRT_arch5.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRT_arch5.png
--------------------------------------------------------------------------------
/cls/figures/HRT_layer6.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRT_layer6.pdf
--------------------------------------------------------------------------------
/cls/figures/HRT_layer6.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/figures/HRT_layer6.png
--------------------------------------------------------------------------------
/cls/install.sh:
--------------------------------------------------------------------------------
1 | git clone https://github.com/NVIDIA/apex
2 | cd apex
3 | /data/anaconda/envs/pytorch1.7.1/bin/python -m pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
--------------------------------------------------------------------------------
/cls/logger.py:
--------------------------------------------------------------------------------
1 | # --------------------------------------------------------
2 | # High Resolution Transformer
3 | # Copyright (c) 2021 Microsoft
4 | # Licensed under The MIT License [see LICENSE for details]
5 | # Written by Ze Liu
6 | # Modified by Rao Fu, RainbowSecret
7 | # --------------------------------------------------------
8 |
9 | import os
10 | import sys
11 | import logging
12 | import functools
13 | from termcolor import colored
14 |
15 |
16 | @functools.lru_cache()
17 | def create_logger(output_dir, dist_rank=0, name=""):
18 | # create logger
19 | logger = logging.getLogger(name)
20 | logger.setLevel(logging.DEBUG)
21 | logger.propagate = False
22 |
23 | # create formatter
24 | fmt = "[%(asctime)s %(name)s] (%(filename)s %(lineno)d): %(levelname)s %(message)s"
25 | color_fmt = (
26 | colored("[%(asctime)s %(name)s]", "green")
27 | + colored("(%(filename)s %(lineno)d)", "yellow")
28 | + ": %(levelname)s %(message)s"
29 | )
30 |
31 | # create console handlers for master process
32 | if dist_rank == 0:
33 | console_handler = logging.StreamHandler(sys.stdout)
34 | console_handler.setLevel(logging.DEBUG)
35 | console_handler.setFormatter(
36 | logging.Formatter(fmt=color_fmt, datefmt="%Y-%m-%d %H:%M:%S")
37 | )
38 | logger.addHandler(console_handler)
39 |
40 | # create file handlers
41 | file_handler = logging.FileHandler(
42 | os.path.join(output_dir, f"log_rank{dist_rank}.txt"), mode="a"
43 | )
44 | file_handler.setLevel(logging.DEBUG)
45 | file_handler.setFormatter(logging.Formatter(fmt=fmt, datefmt="%Y-%m-%d %H:%M:%S"))
46 | logger.addHandler(file_handler)
47 |
48 | return logger
49 |
--------------------------------------------------------------------------------
/cls/models/__init__.py:
--------------------------------------------------------------------------------
1 | from .build import build_model
--------------------------------------------------------------------------------
/cls/models/modules/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/models/modules/__init__.py
--------------------------------------------------------------------------------
/cls/run_dist.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 | MODEL=$1
3 | DATAPATH=$2
4 |
5 | CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python -m torch.distributed.launch \
6 | --nproc_per_node=8 \
7 | --master_port 12345 \
8 | main.py \
9 | --cfg configs/$MODEL.yaml \
10 | --batch-size 128 \
11 | --zip \
12 | --data-path $DATAPATH \
13 | --output_dir "output/$MODEL" \
14 | --tag $(date +%F_%H-%M-%S)
--------------------------------------------------------------------------------
/cls/run_eval.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 | MODEL=$1
3 | RESUME=$2
4 | DATAPATH=$3
5 |
6 | CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.launch \
7 | --nproc_per_node=4 \
8 | --master_port 12345 \
9 | main.py \
10 | --eval \
11 | --zip \
12 | --cfg configs/$MODEL.yaml \
13 | --resume $RESUME \
14 | --data-path $DATAPATH
15 |
--------------------------------------------------------------------------------
/cls/run_flops.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | MODEL=$1
4 | DATAPATH=$2
5 | DATASET=$3
6 |
7 | CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch \
8 | --nproc_per_node=1 \
9 | --master_port 12348 \
10 | test_flops.py \
11 | --mode flops \
12 | --data-path $DATAPATH \
13 | --fig_num 100 \
14 | --data-set $DATASET \
15 | --cfg configs/$MODEL.yaml \
16 | --amp-opt-level O1
17 |
--------------------------------------------------------------------------------
/cls/tools/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/cls/tools/__init__.py
--------------------------------------------------------------------------------
/pose/.github/ISSUE_TEMPLATE/config.yml:
--------------------------------------------------------------------------------
1 | blank_issues_enabled: false
2 |
3 | contact_links:
4 | - name: Common Issues
5 | url: https://mmpose.readthedocs.io/en/latest/faq.html
6 | about: Check if your issue already has solutions
7 | - name: MMPose Documentation
8 | url: https://mmpose.readthedocs.io/en/latest/
9 | about: Check if your question is answered in docs
10 |
--------------------------------------------------------------------------------
/pose/.github/ISSUE_TEMPLATE/error-report.md:
--------------------------------------------------------------------------------
1 | ---
2 | name: Error report
3 | about: Create a report to help us improve
4 | title: ''
5 | labels: ''
6 | assignees: ''
7 |
8 | ---
9 |
10 | Thanks for your error report and we appreciate it a lot.
11 | If you feel we have helped you, give us a STAR! :satisfied:
12 |
13 | **Checklist**
14 |
15 | 1. I have searched related issues but cannot get the expected help.
16 | 1. The bug has not been fixed in the latest version.
17 |
18 | **Describe the bug**
19 |
20 | A clear and concise description of what the bug is.
21 |
22 | **Reproduction**
23 |
24 | - What command or script did you run?
25 |
26 | ```
27 | A placeholder for the command.
28 | ```
29 |
30 | - What config did you run?
31 |
32 | ```
33 | A placeholder for the config.
34 | ```
35 |
36 | - Did you make any modifications on the code or config? Did you understand what you have modified?
37 | - What dataset did you use?
38 |
39 | **Environment**
40 |
41 | 1. Please run `PYTHONPATH=${PWD}:$PYTHONPATH python mmpose/utils/collect_env.py` to collect necessary environment information and paste it here.
42 | 1. You may add addition that may be helpful for locating the problem, such as
43 | - How you installed PyTorch [e.g., pip, conda, source]
44 | - Other environment variables that may be related (such as `$PATH`, `$LD_LIBRARY_PATH`, `$PYTHONPATH`, etc.)
45 |
46 | **Error traceback**
47 |
48 | If applicable, paste the error traceback here.
49 |
50 | ```
51 | A placeholder for traceback.
52 | ```
53 |
54 | **Bug fix**
55 |
56 | If you have already identified the reason, you can provide the information here. If you are willing to create a PR to fix it, please also leave a comment here and that would be much appreciated!
57 |
--------------------------------------------------------------------------------
/pose/.github/ISSUE_TEMPLATE/feature_request.md:
--------------------------------------------------------------------------------
1 | ---
2 | name: Feature request
3 | about: Suggest an idea for this project
4 | title: ''
5 | labels: ''
6 | assignees: ''
7 |
8 | ---
9 |
10 | Thanks for your feature request and we will review and plan for it when necessary.
11 | If you feel we have helped you, give us a STAR! :satisfied:
12 |
13 | **Steps**
14 |
15 | 1. Check if the feature has been requested in the [meta issue](https://github.com/open-mmlab/mmpose/issues/9), and if so, click thumb up button.
16 | 1. Post the feature request in the [meta issue](https://github.com/open-mmlab/mmpose/issues/9), if it is new.
17 |
18 | **Describe the feature**
19 |
20 | **Motivation**
21 |
22 | A clear and concise description of the motivation of the feature.
23 |
24 | 1. Ex1. It is inconvenient when [....].
25 | 1. Ex2. There is a recent paper [....], which is very helpful for [....].
26 |
27 | **Related resources**
28 |
29 | If there is an official code released or third-party implementations, please also provide the information here, which would be very helpful.
30 |
31 | **Additional context**
32 |
33 | Add any other context or screenshots about the feature request here.
34 | If you would like to implement the feature and create a PR, please leave a comment here and that would be much appreciated.
35 |
--------------------------------------------------------------------------------
/pose/.github/ISSUE_TEMPLATE/general_questions.md:
--------------------------------------------------------------------------------
1 | ---
2 | name: General questions
3 | about: Ask general questions to get help
4 | title: ''
5 | labels: ''
6 | assignees: ''
7 |
8 | ---
9 |
--------------------------------------------------------------------------------
/pose/.github/workflows/deploy.yml:
--------------------------------------------------------------------------------
1 | name: deploy
2 |
3 | on: push
4 |
5 | jobs:
6 | build-n-publish:
7 | runs-on: ubuntu-latest
8 | if: startsWith(github.event.ref, 'refs/tags')
9 | steps:
10 | - uses: actions/checkout@v2
11 | - name: Set up Python 3.7
12 | uses: actions/setup-python@v2
13 | with:
14 | python-version: 3.7
15 | - name: Build MMPose
16 | run: |
17 | pip install wheel
18 | python setup.py sdist bdist_wheel
19 | - name: Publish distribution to PyPI
20 | run: |
21 | pip install twine
22 | twine upload dist/* -u __token__ -p ${{ secrets.pypi_password }}
23 |
--------------------------------------------------------------------------------
/pose/.pre-commit-config.yaml:
--------------------------------------------------------------------------------
1 | exclude: ^tests/data/
2 | repos:
3 | - repo: https://gitlab.com/pycqa/flake8
4 | rev: 3.8.3
5 | hooks:
6 | - id: flake8
7 | - repo: https://github.com/asottile/seed-isort-config
8 | rev: v2.2.0
9 | hooks:
10 | - id: seed-isort-config
11 | - repo: https://github.com/timothycrosley/isort
12 | rev: 4.3.21
13 | hooks:
14 | - id: isort
15 | - repo: https://github.com/pre-commit/mirrors-yapf
16 | rev: v0.30.0
17 | hooks:
18 | - id: yapf
19 | - repo: https://github.com/pre-commit/pre-commit-hooks
20 | rev: v3.1.0
21 | hooks:
22 | - id: trailing-whitespace
23 | - id: check-yaml
24 | - id: end-of-file-fixer
25 | - id: requirements-txt-fixer
26 | - id: double-quote-string-fixer
27 | - id: check-merge-conflict
28 | - id: fix-encoding-pragma
29 | args: ["--remove"]
30 | - id: mixed-line-ending
31 | args: ["--fix=lf"]
32 | - repo: https://github.com/jumanjihouse/pre-commit-hooks
33 | rev: 2.1.4
34 | hooks:
35 | - id: markdownlint
36 | args: [ "-r", "~MD002,~MD007,~MD013,~MD024,~MD029,~MD033,~MD034,~MD036" ]
37 | - repo: https://github.com/myint/docformatter
38 | rev: v1.3.1
39 | hooks:
40 | - id: docformatter
41 | args: ["--in-place", "--wrap-descriptions", "79"]
42 |
--------------------------------------------------------------------------------
/pose/.readthedocs.yml:
--------------------------------------------------------------------------------
1 | version: 2
2 |
3 | python:
4 | version: 3.7
5 | install:
6 | - requirements: requirements/docs.txt
7 | - requirements: requirements/readthedocs.txt
8 |
--------------------------------------------------------------------------------
/pose/configs/animal/README.md:
--------------------------------------------------------------------------------
1 | # 2D Animal Keypoint Detection
2 |
3 | 2D animal keypoint detection (animal pose estimation) aims to detect the key-point of different species, including rats,
4 | dogs, macaques, and cheetah. It provides detailed behavioral analysis for neuroscience, medical and ecology applications.
5 |
6 | ## Data preparation
7 |
8 | Please follow [DATA Preparation](/docs/tasks/2d_animal_keypoint.md) to prepare data.
9 |
10 | ## Demo
11 |
12 | Please follow [DEMO](/demo/2d_animal_demo.md) to generate fancy demos.
13 |
14 | 
15 |
16 | 
17 |
18 | 
19 |
--------------------------------------------------------------------------------
/pose/configs/body3d/simple_baseline/README.md:
--------------------------------------------------------------------------------
1 | # A simple yet effective baseline for 3d human pose estimation
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @inproceedings{martinez_2017_3dbaseline,
9 | title={A simple yet effective baseline for 3d human pose estimation},
10 | author={Martinez, Julieta and Hossain, Rayat and Romero, Javier and Little, James J.},
11 | booktitle={ICCV},
12 | year={2017}
13 | }
14 | ```
15 |
16 | ## Results and models
17 |
18 | ### 3D Human Pose Estimation
19 |
20 | #### Results on Human3.6M dataset with ground truth 2D detections
21 |
22 | | Arch | MPJPE | P-MPJPE | ckpt | log |
23 | | :--- | :---: | :---: | :---: | :---: |
24 | | [simple3Dbaseline1](/configs/body3d/simple_baseline/h36m/simple3Dbaseline_h36m.py) | 43.4 | 34.3 | [ckpt](https://download.openmmlab.com/mmpose/body3d/simple_baseline/simple3Dbaseline_h36m-f0ad73a4_20210419.pth) | [log](https://download.openmmlab.com/mmpose/body3d/simple_baseline/20210415_065056.log.json) |
25 |
26 | 1 Differing from the original paper, we didn't apply the `max-norm constraint` because we found this led to a better convergence and performance.
27 |
--------------------------------------------------------------------------------
/pose/configs/bottom_up/README.md:
--------------------------------------------------------------------------------
1 | # 2D Bottom-up Human Body Pose Estimation
2 |
3 | Multi-person human pose estimation is defined as the task of detecting the poses (or keypoints) of all people from an input image.
4 | Bottom-up approaches first detect all the keypoints and then group/associate them into person instances.
5 |
6 | ## Data preparation
7 |
8 | Please follow [DATA Preparation](/docs/tasks/2d_body_keypoint.md) to prepare data.
9 |
10 | ## Demo
11 |
12 | Please follow [Demo](/demo/2d_human_pose_demo.md#2d-human-pose-demo) to run demos.
13 |
14 | 
15 |
--------------------------------------------------------------------------------
/pose/configs/face/README.md:
--------------------------------------------------------------------------------
1 | # 2D Face Landmark Detection
2 |
3 | 2D face landmark detection (also referred to as face alignment) is defined as the task of detecting the face keypoints from an input image.
4 |
5 | Normally, the input images are cropped face images, where the face locates at the center;
6 | or the rough location (or the bounding box) of the hand is provided.
7 |
8 | ## Data preparation
9 |
10 | Please follow [DATA Preparation](/docs/tasks/2d_face_keypoint.md) to prepare data.
11 |
12 | ## Demo
13 |
14 | Please follow [Demo](/demo/2d_face_demo.md) to run demos.
15 |
16 | 
17 |
--------------------------------------------------------------------------------
/pose/configs/face/resnet/README.md:
--------------------------------------------------------------------------------
1 | # Simple baselines for human pose estimation and tracking
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @inproceedings{xiao2018simple,
9 | title={Simple baselines for human pose estimation and tracking},
10 | author={Xiao, Bin and Wu, Haiping and Wei, Yichen},
11 | booktitle={Proceedings of the European conference on computer vision (ECCV)},
12 | pages={466--481},
13 | year={2018}
14 | }
15 | ```
16 |
--------------------------------------------------------------------------------
/pose/configs/fashion/README.md:
--------------------------------------------------------------------------------
1 | # 2D Fashion Landmark Detection
2 |
3 | 2D fashion landmark detection (also referred to as fashion alignment) aims to detect the key-point located at the functional region of clothes, for example the neckline and the cuff.
4 |
5 | ## Data preparation
6 |
7 | Please follow [DATA Preparation](/docs/tasks/2d_fashion_landmark.md) to prepare data.
8 |
--------------------------------------------------------------------------------
/pose/configs/fashion/resnet/README.md:
--------------------------------------------------------------------------------
1 | # Simple baselines for human pose estimation and tracking
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @inproceedings{xiao2018simple,
9 | title={Simple baselines for human pose estimation and tracking},
10 | author={Xiao, Bin and Wu, Haiping and Wei, Yichen},
11 | booktitle={Proceedings of the European conference on computer vision (ECCV)},
12 | pages={466--481},
13 | year={2018}
14 | }
15 | ```
16 |
17 | ## Results and models
18 |
19 | ### 2d Fashion Landmark Detection
20 |
21 | #### Results on DeepFashion val set
22 |
23 | |Set | Arch | Input Size | PCK@0.2 | AUC | EPE | ckpt | log |
24 | | :--- | :---: | :--------: | :------: | :------: | :------: |:------: |:------: |
25 | |upper | [pose_resnet_50](/configs/fashion/resnet/deepfashion/res50_deepfashion_upper_256x192.py) | 256x256 | 0.954 | 0.578 | 16.8 | [ckpt](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_upper_256x192-41794f03_20210124.pth) | [log](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_upper_256x192_20210124.log.json) |
26 | |lower | [pose_resnet_50](/configs/fashion/resnet/deepfashion/res50_deepfashion_lower_256x192.py) | 256x256 | 0.965 | 0.744 | 10.5 | [ckpt](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_lower_256x192-1292a839_20210124.pth) | [log](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_lower_256x192_20210124.log.json) |
27 | |full | [pose_resnet_50](/configs/fashion/resnet/deepfashion/res50_deepfashion_full_256x192.py) | 256x256 | 0.977 | 0.664 | 12.7 | [ckpt](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_full_256x192-0dbd6e42_20210124.pth) | [log](https://download.openmmlab.com/mmpose/fashion/resnet/res50_deepfashion_full_256x192_20210124.log.json) |
28 |
--------------------------------------------------------------------------------
/pose/configs/hand/README.md:
--------------------------------------------------------------------------------
1 | # 2D Hand Pose Estimation
2 |
3 | 2D hand pose estimation is defined as the task of detecting the poses (or keypoints) of the hand from an input image.
4 |
5 | Normally, the input images are cropped hand images, where the hand locates at the center;
6 | or the rough location (or the bounding box) of the hand is provided.
7 |
8 | ## Data preparation
9 |
10 | Please follow [DATA Preparation](/docs/tasks/2d_hand_keypoint.md) to prepare data.
11 |
12 | ## Demo
13 |
14 | Please follow [Demo](/demo/2d_hand_demo.md) to run demos.
15 |
16 | 
17 |
--------------------------------------------------------------------------------
/pose/configs/mesh/hmr/README.md:
--------------------------------------------------------------------------------
1 | # End-to-end Recovery of Human Shape and Pose
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @inProceedings{kanazawaHMR18,
9 | title={End-to-end Recovery of Human Shape and Pose},
10 | author = {Angjoo Kanazawa
11 | and Michael J. Black
12 | and David W. Jacobs
13 | and Jitendra Malik},
14 | booktitle={Computer Vision and Pattern Recognition (CVPR)},
15 | year={2018}
16 | }
17 | ```
18 |
19 | ## Results and models
20 |
21 | ### 3D Human Mesh Estimation
22 |
23 | #### Results on Human3.6M with ground-truth bounding box having MPJPE-PA of 52.60 mm on Protocol2
24 |
25 | | Arch | Input Size | MPJPE (P1)| MPJPE-PA (P1) | MPJPE (P2) | MPJPE-PA (P2) | ckpt | log |
26 | | :-------------- | :-----------: | :------: | :------: | :------: | :------: | :------: |:------: |
27 | | [hmr_resnet_50](/configs/mesh/hmr/hmr_res50_224x224.py) | 224x224 | 80.75 | 55.08 | 80.35 | 52.60 | [ckpt](https://download.openmmlab.com/mmpose/mesh/hmr/hmr_mesh_224x224-c21e8229_20201015.pth) | [log](https://download.openmmlab.com/mmpose/mesh/hmr/hmr_mesh_224x224_20201015.log.json) |
28 |
--------------------------------------------------------------------------------
/pose/configs/top_down/README.md:
--------------------------------------------------------------------------------
1 | # 2D Top-down Human Body Pose Estimation
2 |
3 | Multi-person human pose estimation is defined as the task of detecting the poses (or keypoints) of all people from an input image.
4 | Top-down methods divide the task into two stages: human detection and pose estimation.
5 | They perform human detection first, followed by single-person pose estimation given human bounding boxes.
6 |
7 | ## Data preparation
8 |
9 | Please follow [DATA Preparation](/docs/tasks/2d_body_keypoint.md) to prepare data.
10 |
11 | ## Demo
12 |
13 | Please follow [Demo](/demo/2d_human_pose_demo.md#2d-human-pose-demo) to run demos.
14 |
15 |
16 |
--------------------------------------------------------------------------------
/pose/configs/top_down/alexnet/README.md:
--------------------------------------------------------------------------------
1 | # Imagenet classification with deep convolutional neural networks
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @inproceedings{krizhevsky2012imagenet,
9 | title={Imagenet classification with deep convolutional neural networks},
10 | author={Krizhevsky, Alex and Sutskever, Ilya and Hinton, Geoffrey E},
11 | booktitle={Advances in neural information processing systems},
12 | pages={1097--1105},
13 | year={2012}
14 | }
15 | ```
16 |
17 | ## Results and models
18 |
19 | ### 2d Human Pose Estimation
20 |
21 | #### Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
22 |
23 | | Arch | Input Size | AP | AP50 | AP75 | AR | AR50 | ckpt | log |
24 | | :----------------- | :-----------: | :------: | :------: | :------: | :------: | :------: |:------: |:------: |
25 | | [pose_alexnet](/configs/top_down/alexnet/coco/alexnet_coco_256x192.py) | 256x192 | 0.397 | 0.758 | 0.381 | 0.478 | 0.822 | [ckpt](https://download.openmmlab.com/mmpose/top_down/alexnet/alexnet_coco_256x192-a7b1fd15_20200727.pth) | [log](https://download.openmmlab.com/mmpose/top_down/alexnet/alexnet_coco_256x192_20200727.log.json) |
26 |
--------------------------------------------------------------------------------
/pose/configs/top_down/vgg/README.md:
--------------------------------------------------------------------------------
1 | # Very Deep Convolutional Networks for Large-Scale Image Recognition
2 |
3 | ## Introduction
4 |
5 |
6 |
7 | ```bibtex
8 | @article{simonyan2014very,
9 | title={Very deep convolutional networks for large-scale image recognition},
10 | author={Simonyan, Karen and Zisserman, Andrew},
11 | journal={arXiv preprint arXiv:1409.1556},
12 | year={2014}
13 | }
14 |
15 | ```
16 |
--------------------------------------------------------------------------------
/pose/configs/wholebody/README.md:
--------------------------------------------------------------------------------
1 | # 2D Human Whole-Body Pose Estimation (Top-down)
2 |
3 | 2D human whole-body pose estimation aims to localize dense landmarks on the entire human body including face, hands, body, and feet.
4 | Top-down methods divide the task into two stages: human detection and whole-body pose estimation.
5 | They perform human detection first, followed by single-person whole-body pose estimation given human bounding boxes.
6 |
7 | ## Data preparation
8 |
9 | Please follow [DATA Preparation](/docs/tasks/2d_wholebody_keypoint.md) to prepare data.
10 |
11 | ## Demo
12 |
13 | Please follow [Demo](/demo/2d_wholebody_pose_demo.md) to run demos.
14 |
15 | 
16 |
--------------------------------------------------------------------------------
/pose/demo/README.md:
--------------------------------------------------------------------------------
1 | # Demo
2 |
3 | This page provides tutorials about running demos. Please click the caption for more information.
4 |
5 |
6 |
7 |
8 |
9 |
10 | [2D human pose demo](docs/2d_human_pose_demo.md)
11 |
14 |
15 |
16 |
17 |
18 | [2D human whole-body pose demo](docs/2d_wholebody_pose_demo.md)
19 |
22 |
23 |
24 |
25 |
26 | [2D hand pose demo](docs/2d_hand_demo.md)
27 |
30 |
31 |
32 |
33 |
34 | [2D face keypoint demo](docs/2d_face_demo.md)
35 |
38 |
39 |
40 |
41 |
42 | [2D pose tracking demo](docs/2d_pose_tracking_demo.md)
43 |
46 |
47 |
48 | [2D animal_pose demo](docs/2d_animal_demo.md)
49 |
2 |
3 | int deform_conv_forward(THFloatTensor *input, THFloatTensor *offset,
4 | THFloatTensor *output)
5 | {
6 | // if (!THFloatTensor_isSameSizeAs(input1, input2))
7 | // return 0;
8 | // THFloatTensor_resizeAs(output, input);
9 | // THFloatTensor_cadd(output, input1, 1.0, input2);
10 | return 1;
11 | }
12 |
13 | int deform_conv_backward(THFloatTensor *grad_output, THFloatTensor *grad_input,
14 | THFloatTensor *grad_offset)
15 | {
16 | // THFloatTensor_resizeAs(grad_input, grad_output);
17 | // THFloatTensor_fill(grad_input, 1);
18 | return 1;
19 | }
20 |
--------------------------------------------------------------------------------
/seg/lib/extensions/dcn/src/deform_conv.h:
--------------------------------------------------------------------------------
1 | int deform_conv_forward(THFloatTensor *input, THFloatTensor *offset,
2 | THFloatTensor *output);
3 | int deform_conv_backward(THFloatTensor *grad_output, THFloatTensor *grad_input,
4 | THFloatTensor *grad_offset);
5 |
--------------------------------------------------------------------------------
/seg/lib/extensions/dcn/src/deform_conv_cuda.h:
--------------------------------------------------------------------------------
1 | int deform_conv_forward_cuda(THCudaTensor *input,
2 | THCudaTensor *weight, /*THCudaTensor * bias, */
3 | THCudaTensor *offset, THCudaTensor *output,
4 | THCudaTensor *columns, THCudaTensor *ones, int kW,
5 | int kH, int dW, int dH, int padW, int padH,
6 | int dilationW, int dilationH,
7 | int deformable_group, int im2col_step);
8 |
9 | int deform_conv_backward_input_cuda(
10 | THCudaTensor *input, THCudaTensor *offset, THCudaTensor *gradOutput,
11 | THCudaTensor *gradInput, THCudaTensor *gradOffset, THCudaTensor *weight,
12 | THCudaTensor *columns, int kW, int kH, int dW, int dH, int padW, int padH,
13 | int dilationW, int dilationH, int deformable_group, int im2col_step);
14 |
15 | int deform_conv_backward_parameters_cuda(
16 | THCudaTensor *input, THCudaTensor *offset, THCudaTensor *gradOutput,
17 | THCudaTensor *gradWeight, /*THCudaTensor *gradBias, */
18 | THCudaTensor *columns, THCudaTensor *ones, int kW, int kH, int dW, int dH,
19 | int padW, int padH, int dilationW, int dilationH, int deformable_group,
20 | float scale, int im2col_step);
21 |
--------------------------------------------------------------------------------
/seg/lib/extensions/dcn/src/modulated_dcn.c:
--------------------------------------------------------------------------------
1 | #include |
2 | #include
3 | #include
4 |
5 | void modulated_deform_conv_forward(THFloatTensor *input, THFloatTensor *weight,
6 | THFloatTensor *bias, THFloatTensor *ones,
7 | THFloatTensor *offset, THFloatTensor *mask,
8 | THFloatTensor *output, THFloatTensor *columns,
9 | const int pad_h, const int pad_w,
10 | const int stride_h, const int stride_w,
11 | const int dilation_h, const int dilation_w,
12 | const int deformable_group)
13 | {
14 | printf("only implemented in GPU");
15 | }
16 | void modulated_deform_conv_backward(THFloatTensor *input, THFloatTensor *weight,
17 | THFloatTensor *bias, THFloatTensor *ones,
18 | THFloatTensor *offset, THFloatTensor *mask,
19 | THFloatTensor *output, THFloatTensor *columns,
20 | THFloatTensor *grad_input, THFloatTensor *grad_weight,
21 | THFloatTensor *grad_bias, THFloatTensor *grad_offset,
22 | THFloatTensor *grad_mask, THFloatTensor *grad_output,
23 | int kernel_h, int kernel_w,
24 | int stride_h, int stride_w,
25 | int pad_h, int pad_w,
26 | int dilation_h, int dilation_w,
27 | int deformable_group)
28 | {
29 | printf("only implemented in GPU");
30 | }
--------------------------------------------------------------------------------
/seg/lib/extensions/dcn/src/modulated_dcn.h:
--------------------------------------------------------------------------------
1 | void modulated_deform_conv_forward(THFloatTensor *input, THFloatTensor *weight,
2 | THFloatTensor *bias, THFloatTensor *ones,
3 | THFloatTensor *offset, THFloatTensor *mask,
4 | THFloatTensor *output, THFloatTensor *columns,
5 | const int pad_h, const int pad_w,
6 | const int stride_h, const int stride_w,
7 | const int dilation_h, const int dilation_w,
8 | const int deformable_group);
9 | void modulated_deform_conv_backward(THFloatTensor *input, THFloatTensor *weight,
10 | THFloatTensor *bias, THFloatTensor *ones,
11 | THFloatTensor *offset, THFloatTensor *mask,
12 | THFloatTensor *output, THFloatTensor *columns,
13 | THFloatTensor *grad_input, THFloatTensor *grad_weight,
14 | THFloatTensor *grad_bias, THFloatTensor *grad_offset,
15 | THFloatTensor *grad_mask, THFloatTensor *grad_output,
16 | int kernel_h, int kernel_w,
17 | int stride_h, int stride_w,
18 | int pad_h, int pad_w,
19 | int dilation_h, int dilation_w,
20 | int deformable_group);
--------------------------------------------------------------------------------
/seg/lib/extensions/dcn/test.py:
--------------------------------------------------------------------------------
1 | import torch
2 | import torch.nn as nn
3 | import torch.nn.functional as F
4 | from torch.autograd import Variable
5 |
6 | from modules import DeformConv
7 |
8 | num_deformable_groups = 2
9 |
10 | N, inC, inH, inW = 2, 6, 512, 512
11 | outC, outH, outW = 4, 512, 512
12 | kH, kW = 3, 3
13 |
14 | conv = nn.Conv2d(
15 | inC,
16 | num_deformable_groups * 2 * kH * kW,
17 | kernel_size=(kH, kW),
18 | stride=(1, 1),
19 | padding=(1, 1),
20 | bias=False).cuda()
21 |
22 | conv_offset2d = DeformConv(
23 | inC,
24 | outC, (kH, kW),
25 | stride=1,
26 | padding=1,
27 | num_deformable_groups=num_deformable_groups).cuda()
28 |
29 | inputs = Variable(torch.randn(N, inC, inH, inW).cuda(), requires_grad=True)
30 | offset = conv(inputs)
31 | #offset = Variable(torch.randn(N, num_deformable_groups * 2 * kH * kW, inH, inW).cuda(), requires_grad=True)
32 | output = conv_offset2d(inputs, offset)
33 | output.backward(output.data)
34 | print(output.size())
35 |
--------------------------------------------------------------------------------
/seg/lib/extensions/frn/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/extensions/frn/__init__.py
--------------------------------------------------------------------------------
/seg/lib/extensions/frn/frn.py:
--------------------------------------------------------------------------------
1 | import torch
2 | import torch.nn as nn
3 |
4 | class FilterResponseNormalization(nn.Module):
5 | def __init__(self, beta, gamma, tau, eps=1e-6):
6 | """
7 | Input Variables:
8 | ----------------
9 | beta, gamma, tau: Variables of shape [1, C, 1, 1].
10 | eps: A scalar constant or learnable variable.
11 | """
12 |
13 | super(FilterResponseNormalization, self).__init__()
14 | self.beta = beta
15 | self.gamma = gamma
16 | self.tau = tau
17 | self.eps = torch.Tensor([eps])
18 |
19 | def forward(self, x):
20 | """
21 | Input Variables:
22 | ----------------
23 | x: Input tensor of shape [NxCxHxW]
24 | """
25 |
26 | n, c, h, w = x.shape
27 | assert (self.gamma.shape[1], self.beta.shape[1], self.tau.shape[1]) == (c, c, c)
28 |
29 | # Compute the mean norm of activations per channel
30 | nu2 = torch.mean(x.pow(2), (2,3), keepdims=True)
31 | # Perform FRN
32 | x = x * torch.rsqrt(nu2 + torch.abs(self.eps))
33 | # Return after applying the Offset-ReLU non-linearity
34 | return torch.max(self.gamma*x + self.beta, self.tau)
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/extensions/inplace_abn/__init__.py
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn/src/inplace_abn.h:
--------------------------------------------------------------------------------
1 | #pragma once
2 |
3 | #include
4 |
5 | #include
6 |
7 | std::vector mean_var_cpu(at::Tensor x);
8 | std::vector mean_var_cuda(at::Tensor x);
9 |
10 | at::Tensor forward_cpu(at::Tensor x, at::Tensor mean, at::Tensor var, at::Tensor weight, at::Tensor bias,
11 | bool affine, float eps);
12 | at::Tensor forward_cuda(at::Tensor x, at::Tensor mean, at::Tensor var, at::Tensor weight, at::Tensor bias,
13 | bool affine, float eps);
14 |
15 | std::vector edz_eydz_cpu(at::Tensor z, at::Tensor dz, at::Tensor weight, at::Tensor bias,
16 | bool affine, float eps);
17 | std::vector edz_eydz_cuda(at::Tensor z, at::Tensor dz, at::Tensor weight, at::Tensor bias,
18 | bool affine, float eps);
19 |
20 | std::vector backward_cpu(at::Tensor z, at::Tensor dz, at::Tensor var, at::Tensor weight, at::Tensor bias,
21 | at::Tensor edz, at::Tensor eydz, bool affine, float eps);
22 | std::vector backward_cuda(at::Tensor z, at::Tensor dz, at::Tensor var, at::Tensor weight, at::Tensor bias,
23 | at::Tensor edz, at::Tensor eydz, bool affine, float eps);
24 |
25 | void leaky_relu_backward_cpu(at::Tensor z, at::Tensor dz, float slope);
26 | void leaky_relu_backward_cuda(at::Tensor z, at::Tensor dz, float slope);
27 |
28 | void elu_backward_cpu(at::Tensor z, at::Tensor dz);
29 | void elu_backward_cuda(at::Tensor z, at::Tensor dz);
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn_1/__init__.py:
--------------------------------------------------------------------------------
1 | from .bn import ABN, InPlaceABN, InPlaceABNSync
2 | from .functions import ACT_RELU, ACT_LEAKY_RELU, ACT_ELU, ACT_NONE
3 |
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn_1/misc.py:
--------------------------------------------------------------------------------
1 | import torch.nn as nn
2 | import torch
3 | import torch.distributed as dist
4 |
5 | class GlobalAvgPool2d(nn.Module):
6 | def __init__(self):
7 | """Global average pooling over the input's spatial dimensions"""
8 | super(GlobalAvgPool2d, self).__init__()
9 |
10 | def forward(self, inputs):
11 | in_size = inputs.size()
12 | return inputs.view((in_size[0], in_size[1], -1)).mean(dim=2)
13 |
14 | class SingleGPU(nn.Module):
15 | def __init__(self, module):
16 | super(SingleGPU, self).__init__()
17 | self.module=module
18 |
19 | def forward(self, input):
20 | return self.module(input.cuda(non_blocking=True))
21 |
22 |
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn_1/src/checks.h:
--------------------------------------------------------------------------------
1 | #pragma once
2 |
3 | #include
4 |
5 | // Define AT_CHECK for old version of ATen where the same function was called AT_ASSERT
6 | #ifndef AT_CHECK
7 | #define AT_CHECK AT_ASSERT
8 | #endif
9 |
10 | #define CHECK_CUDA(x) AT_CHECK((x).type().is_cuda(), #x " must be a CUDA tensor")
11 | #define CHECK_CPU(x) AT_CHECK(!(x).type().is_cuda(), #x " must be a CPU tensor")
12 | #define CHECK_CONTIGUOUS(x) AT_CHECK((x).is_contiguous(), #x " must be contiguous")
13 |
14 | #define CHECK_CUDA_INPUT(x) CHECK_CUDA(x); CHECK_CONTIGUOUS(x)
15 | #define CHECK_CPU_INPUT(x) CHECK_CPU(x); CHECK_CONTIGUOUS(x)
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn_1/src/utils/checks.h:
--------------------------------------------------------------------------------
1 | #pragma once
2 |
3 | #include
4 |
5 | // Define AT_CHECK for old version of ATen where the same function was called AT_ASSERT
6 | #ifndef AT_CHECK
7 | #define AT_CHECK AT_ASSERT
8 | #endif
9 |
10 | #define CHECK_CUDA(x) AT_CHECK((x).type().is_cuda(), #x " must be a CUDA tensor")
11 | #define CHECK_CPU(x) AT_CHECK(!(x).type().is_cuda(), #x " must be a CPU tensor")
12 | #define CHECK_CONTIGUOUS(x) AT_CHECK((x).is_contiguous(), #x " must be contiguous")
13 |
14 | #define CHECK_CUDA_INPUT(x) CHECK_CUDA(x); CHECK_CONTIGUOUS(x)
15 | #define CHECK_CPU_INPUT(x) CHECK_CPU(x); CHECK_CONTIGUOUS(x)
--------------------------------------------------------------------------------
/seg/lib/extensions/inplace_abn_1/src/utils/common.h:
--------------------------------------------------------------------------------
1 | #pragma once
2 |
3 | #include
4 |
5 | /*
6 | * Functions to share code between CPU and GPU
7 | */
8 |
9 | #ifdef __CUDACC__
10 | // CUDA versions
11 |
12 | #define HOST_DEVICE __host__ __device__
13 | #define INLINE_HOST_DEVICE __host__ __device__ inline
14 | #define FLOOR(x) floor(x)
15 |
16 | #if __CUDA_ARCH__ >= 600
17 | // Recent compute capabilities have block-level atomicAdd for all data types, so we use that
18 | #define ACCUM(x,y) atomicAdd_block(&(x),(y))
19 | #else
20 | // Older architectures don't have block-level atomicAdd, nor atomicAdd for doubles, so we defer to atomicAdd for float
21 | // and use the known atomicCAS-based implementation for double
22 | template
23 | __device__ inline data_t atomic_add(data_t *address, data_t val) {
24 | return atomicAdd(address, val);
25 | }
26 |
27 | template<>
28 | __device__ inline double atomic_add(double *address, double val) {
29 | unsigned long long int* address_as_ull = (unsigned long long int*)address;
30 | unsigned long long int old = *address_as_ull, assumed;
31 | do {
32 | assumed = old;
33 | old = atomicCAS(address_as_ull, assumed, __double_as_longlong(val + __longlong_as_double(assumed)));
34 | } while (assumed != old);
35 | return __longlong_as_double(old);
36 | }
37 |
38 | #define ACCUM(x,y) atomic_add(&(x),(y))
39 | #endif // #if __CUDA_ARCH__ >= 600
40 |
41 | #else
42 | // CPU versions
43 |
44 | #define HOST_DEVICE
45 | #define INLINE_HOST_DEVICE inline
46 | #define FLOOR(x) std::floor(x)
47 | #define ACCUM(x,y) (x) += (y)
48 |
49 | #endif // #ifdef __CUDACC__
--------------------------------------------------------------------------------
/seg/lib/extensions/pacnet/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/extensions/pacnet/__init__.py
--------------------------------------------------------------------------------
/seg/lib/extensions/parallel/__init__.py:
--------------------------------------------------------------------------------
1 | from .data_container import DataContainer
2 | from .distributed import MMDistributedDataParallel
3 | from .scatter_gather import scatter, scatter_kwargs
4 |
5 | __all__ = [
6 | 'collate', 'DataContainer', 'MMDataParallel', 'MMDistributedDataParallel',
7 | 'scatter', 'scatter_kwargs'
8 | ]
9 |
--------------------------------------------------------------------------------
/seg/lib/extensions/switchablenorms/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/extensions/switchablenorms/__init__.py
--------------------------------------------------------------------------------
/seg/lib/extensions/syncbn/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/extensions/syncbn/__init__.py
--------------------------------------------------------------------------------
/seg/lib/extensions/syncbn/src/operator.cpp:
--------------------------------------------------------------------------------
1 | #include "operator.h"
2 |
3 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
4 | m.def("batchnorm_forward", &BatchNorm_Forward_CUDA, "BatchNorm forward (CUDA)");
5 | m.def("batchnorm_backward", &BatchNorm_Backward_CUDA, "BatchNorm backward (CUDA)");
6 | m.def("sumsquare_forward", &Sum_Square_Forward_CUDA, "SumSqu forward (CUDA)");
7 | m.def("sumsquare_backward", &Sum_Square_Backward_CUDA, "SumSqu backward (CUDA)");
8 | }
9 |
--------------------------------------------------------------------------------
/seg/lib/extensions/syncbn/src/operator.h:
--------------------------------------------------------------------------------
1 | #include
2 | #include
3 |
4 |
5 | at::Tensor BatchNorm_Forward_CUDA(
6 | const at::Tensor input_,
7 | const at::Tensor mean_,
8 | const at::Tensor std_,
9 | const at::Tensor gamma_,
10 | const at::Tensor beta_);
11 |
12 | std::vector BatchNorm_Backward_CUDA(
13 | const at::Tensor gradoutput_,
14 | const at::Tensor input_,
15 | const at::Tensor mean_,
16 | const at::Tensor std_,
17 | const at::Tensor gamma_,
18 | const at::Tensor beta_,
19 | bool train);
20 |
21 | std::vector Sum_Square_Forward_CUDA(
22 | const at::Tensor input_);
23 |
24 | at::Tensor Sum_Square_Backward_CUDA(
25 | const at::Tensor input_,
26 | const at::Tensor gradSum_,
27 | const at::Tensor gradSquare_);
28 |
--------------------------------------------------------------------------------
/seg/lib/loss/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/loss/__init__.py
--------------------------------------------------------------------------------
/seg/lib/metrics/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/metrics/__init__.py
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/metrics/cityscapes/__init__.py
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/evaluation/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/metrics/cityscapes/evaluation/__init__.py
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/evaluation/addToConfusionMatrix_impl.c:
--------------------------------------------------------------------------------
1 | // cython methods to speed-up evaluation
2 |
3 | void addToConfusionMatrix( const unsigned char* f_prediction_p ,
4 | const unsigned char* f_groundTruth_p ,
5 | const unsigned int f_width_i ,
6 | const unsigned int f_height_i ,
7 | unsigned long long* f_confMatrix_p ,
8 | const unsigned int f_confMatDim_i )
9 | {
10 | const unsigned int size_ui = f_height_i * f_width_i;
11 | for (unsigned int i = 0; i < size_ui; ++i)
12 | {
13 | const unsigned char predPx = f_prediction_p [i];
14 | const unsigned char gtPx = f_groundTruth_p[i];
15 | f_confMatrix_p[f_confMatDim_i*gtPx + predPx] += 1u;
16 | }
17 | }
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/evaluation/instance.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/python
2 | #
3 | # Instance class
4 | #
5 |
6 | class Instance(object):
7 | instID = 0
8 | labelID = 0
9 | pixelCount = 0
10 | medDist = -1
11 | distConf = 0.0
12 |
13 | def __init__(self, imgNp, instID):
14 | if (instID == -1):
15 | return
16 | self.instID = int(instID)
17 | self.labelID = int(self.getLabelID(instID))
18 | self.pixelCount = int(self.getInstancePixels(imgNp, instID))
19 |
20 | def getLabelID(self, instID):
21 | if (instID < 1000):
22 | return instID
23 | else:
24 | return int(instID / 1000)
25 |
26 | def getInstancePixels(self, imgNp, instLabel):
27 | return (imgNp == instLabel).sum()
28 |
29 | def toJSON(self):
30 | return json.dumps(self, default=lambda o: o.__dict__, sort_keys=True, indent=4)
31 |
32 | def toDict(self):
33 | buildDict = {}
34 | buildDict["instID"] = self.instID
35 | buildDict["labelID"] = self.labelID
36 | buildDict["pixelCount"] = self.pixelCount
37 | buildDict["medDist"] = self.medDist
38 | buildDict["distConf"] = self.distConf
39 | return buildDict
40 |
41 | def fromJSON(self, data):
42 | self.instID = int(data["instID"])
43 | self.labelID = int(data["labelID"])
44 | self.pixelCount = int(data["pixelCount"])
45 | if ("medDist" in data):
46 | self.medDist = float(data["medDist"])
47 | self.distConf = float(data["distConf"])
48 |
49 | def __str__(self):
50 | return "("+str(self.instID)+")"
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/helpers/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/metrics/cityscapes/helpers/__init__.py
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/make.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | # check the enviroment info
4 |
5 | # PYTHON="/root/miniconda3/bin/python"
6 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
7 | export PYTHONPATH="/msravcshare/yuyua/code/segmentation/openseg.pytorch":$PYTHONPATH
8 |
9 | cd ../../../
10 | ${PYTHON} lib/metrics/cityscapes/setup.py build_ext --inplace
11 |
--------------------------------------------------------------------------------
/seg/lib/metrics/cityscapes/setup.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/python
2 | #
3 | # Enable cython support for eval metrics
4 | # Run as
5 | # setup.py build_ext --inplace
6 | #
7 | # WARNING: Only tested for Ubuntu 64bit OS.
8 |
9 | try:
10 | from distutils.core import setup
11 | from Cython.Build import cythonize
12 | except:
13 | print("Unable to setup. Please use pip to install: cython")
14 | print("sudo pip install cython")
15 | import os
16 | import numpy
17 |
18 | os.environ["CC"] = "g++"
19 | os.environ["CXX"] = "g++"
20 |
21 | setup(ext_modules = cythonize("lib/metrics/cityscapes/evaluation/addToConfusionMatrix.pyx"),
22 | include_dirs=[numpy.get_include()])
23 |
--------------------------------------------------------------------------------
/seg/lib/models/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/backbones/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/hrnet/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/backbones/hrnet/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/hrt/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/backbones/hrt/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/hrt/modules/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/backbones/hrt/modules/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/resnet/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/backbones/resnet/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/backbones/vit/helper.py:
--------------------------------------------------------------------------------
1 | import torch.nn as nn
2 |
3 |
4 | class IntermediateSequential(nn.Sequential):
5 | def __init__(self, *args, return_intermediate=True):
6 | super().__init__(*args)
7 | self.return_intermediate = return_intermediate
8 |
9 | def forward(self, input):
10 | if not self.return_intermediate:
11 | return super().forward(input)
12 |
13 | intermediate_outputs = {}
14 | output = input
15 | for name, module in self.named_children():
16 | output = intermediate_outputs[name] = module(output)
17 |
18 | return output, intermediate_outputs
--------------------------------------------------------------------------------
/seg/lib/models/backbones/vit/position_encoding.py:
--------------------------------------------------------------------------------
1 | import torch
2 | import torch.nn as nn
3 |
4 |
5 | class FixedPositionalEncoding(nn.Module):
6 | def __init__(self, embedding_dim, max_length=5000):
7 | super(FixedPositionalEncoding, self).__init__()
8 |
9 | pe = torch.zeros(max_length, embedding_dim)
10 | position = torch.arange(0, max_length, dtype=torch.float).unsqueeze(1)
11 | div_term = torch.exp(
12 | torch.arange(0, embedding_dim, 2).float()
13 | * (-torch.log(torch.tensor(10000.0)) / embedding_dim)
14 | )
15 | pe[:, 0::2] = torch.sin(position * div_term)
16 | pe[:, 1::2] = torch.cos(position * div_term)
17 | pe = pe.unsqueeze(0).transpose(0, 1)
18 | self.register_buffer('pe', pe)
19 |
20 | def forward(self, x):
21 | x = x + self.pe[: x.size(0), :]
22 | return x
23 |
24 |
25 | class LearnedPositionalEncoding(nn.Module):
26 | def __init__(self, max_position_embeddings, embedding_dim, seq_length):
27 | super(LearnedPositionalEncoding, self).__init__()
28 | self.pe = nn.Embedding(max_position_embeddings, embedding_dim)
29 | self.seq_length = seq_length
30 |
31 | self.register_buffer(
32 | "position_ids",
33 | torch.arange(max_position_embeddings).expand((1, -1)),
34 | )
35 |
36 | def forward(self, x, position_ids=None):
37 | if position_ids is None:
38 | position_ids = self.position_ids[:, : self.seq_length]
39 |
40 | position_embeddings = self.pe(position_ids)
41 | return x + position_embeddings
--------------------------------------------------------------------------------
/seg/lib/models/modules/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/modules/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/modules/infer_time.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 | SCRIPTPATH="$( cd "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"
3 | cd $SCRIPTPATH
4 | cd ../../../
5 | . config.profile
6 |
7 | export PYTHONPATH="../../../":$PYTHONPATH
8 | export eval_os_8=1
9 | export bn_type="inplace_abn"
10 |
11 | # ${PYTHON} -m lib.models.modules.psp_block
12 | # ${PYTHON} -m lib.models.modules.aspp_block
13 | # ${PYTHON} -m lib.models.modules.base_oc_block
14 | export isa_type="base_oc"
15 | ${PYTHON} -m lib.models.modules.isa_block
16 | ${PYTHON} -m lib.models.modules.spatial_ocr_block
17 |
--------------------------------------------------------------------------------
/seg/lib/models/modules/time_log:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/modules/time_log
--------------------------------------------------------------------------------
/seg/lib/models/nets/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/nets/__init__.py
--------------------------------------------------------------------------------
/seg/lib/models/tools/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/models/tools/__init__.py
--------------------------------------------------------------------------------
/seg/lib/utils/__init__.py:
--------------------------------------------------------------------------------
1 |
2 |
--------------------------------------------------------------------------------
/seg/lib/utils/helpers/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/utils/helpers/__init__.py
--------------------------------------------------------------------------------
/seg/lib/utils/helpers/dc_helper.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env python
2 | # -*- coding:utf-8 -*-
3 | # Author: Donny You(youansheng@gmail.com)
4 |
5 |
6 | from __future__ import absolute_import
7 | from __future__ import division
8 | from __future__ import print_function
9 |
10 | import itertools
11 |
12 | from lib.extensions.parallel.data_container import DataContainer
13 |
14 |
15 | class DCHelper(object):
16 |
17 | @staticmethod
18 | def tolist(dc):
19 | return list(itertools.chain(*dc.data))
20 |
21 | @staticmethod
22 | def todc(data_list, gpu_list, cpu_only=False):
23 | assert len(data_list) % len(gpu_list) == 0
24 | samples_per_gpu = len(data_list) // len(gpu_list)
25 | stacked = []
26 | for i in range(0, len(data_list), samples_per_gpu):
27 | stacked.append(data_list[i:i + samples_per_gpu])
28 |
29 | return DataContainer(stacked, cpu_only=cpu_only)
30 |
--------------------------------------------------------------------------------
/seg/lib/utils/tools/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/utils/tools/__init__.py
--------------------------------------------------------------------------------
/seg/lib/utils/tools/average_meter.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env python
2 | #-*- coding:utf-8 -*-
3 | # Author: Donny You (youansheng@gmail.com)
4 | # Utils to store the average and current value.
5 |
6 |
7 | from __future__ import absolute_import
8 | from __future__ import division
9 | from __future__ import print_function
10 |
11 |
12 | class AverageMeter(object):
13 | """ Computes ans stores the average and current value"""
14 | def __init__(self):
15 | self.reset()
16 |
17 | def reset(self):
18 | self.val = 0.
19 | self.avg = 0.
20 | self.sum = 0.
21 | self.count = 0
22 |
23 | def update(self, val, n=1):
24 | self.val = val
25 | self.sum += val * n
26 | self.count += n
27 | self.avg = self.sum / self.count
28 |
--------------------------------------------------------------------------------
/seg/lib/utils/tools/timer.py:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env python
2 | # -*- coding:utf-8 -*-
3 | # Author: Donny You(youansheng@gmail.com)
4 |
5 |
6 | from __future__ import absolute_import
7 | from __future__ import division
8 | from __future__ import print_function
9 | from __future__ import unicode_literals
10 |
11 | import time
12 |
13 |
14 | class Timer(object):
15 | """A simple timer."""
16 | def __init__(self):
17 | self.reset()
18 |
19 | def tic(self):
20 | # using time.time instead of time.clock because time time.clock
21 | # does not normalize for multithreading
22 | self.start_time = time.time()
23 |
24 | def toc(self, average=True):
25 | self.diff = time.time() - self.start_time
26 | self.total_time += self.diff
27 | self.calls += 1
28 | self.average_time = self.total_time / self.calls
29 | if average:
30 | return self.average_time
31 | else:
32 | return self.diff
33 |
34 | def reset(self):
35 | self.total_time = 0.
36 | self.calls = 0
37 | self.start_time = 0.
38 | self.diff = 0.
39 | self.average_time = 0.
--------------------------------------------------------------------------------
/seg/lib/vis/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/vis/__init__.py
--------------------------------------------------------------------------------
/seg/lib/vis/color150.mat:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/vis/color150.mat
--------------------------------------------------------------------------------
/seg/lib/vis/color60.mat:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/lib/vis/color60.mat
--------------------------------------------------------------------------------
/seg/requirements.txt:
--------------------------------------------------------------------------------
1 | cython
2 | numpy
3 | cffi
4 | opencv-python
5 | scipy
6 | easydict
7 | matplotlib
8 | Pillow>=6.2.2
9 | torchcontrib
10 | yacs
11 | pyyaml
12 | visdom
13 | bs4
14 | html5lib
15 | ninja
16 | torch>=1.7.0
17 | torchvision>=0.6.0
18 | timm
19 | einops
20 | git+https://github.com/lucasb-eyer/pydensecrf.git#egg=pydensecrf
21 | git+https://github.com/hsfzxjy/mocona@v0.1.0#egg=mocona
--------------------------------------------------------------------------------
/seg/run_dist.sh:
--------------------------------------------------------------------------------
1 | CONFIG=$1
2 | MODE=$2
3 | TAG=$3
4 | python -m pip install timm
5 | python -m pip install einops
6 | pip uninstall PIL
7 | pip install pillow
8 | CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash scripts/$CONFIG.sh $MODE $TAG .
9 | # CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash scripts/$CONFIG.sh resume run0_ .
--------------------------------------------------------------------------------
/seg/run_local.sh:
--------------------------------------------------------------------------------
1 |
2 | CUDA_VISIBLE_DEVICES=0,1,2,3 bash scripts/coco_stuff/hrt/run_hrt_small_ocr_v2_ohem.sh local
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrnet_w48_ocr_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | # PYTHON="python"
7 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
8 |
9 | # check the enviroment info
10 | export PYTHONPATH="$PWD":$PYTHONPATH
11 | DATA_DIR="${DATA_ROOT}/cityscapes"
12 | BACKBONE="hrnet48"
13 | MODEL_NAME="hrnet_w48_ocr"
14 | CONFIGS="configs/cityscapes/H_48_D_4.json"
15 |
16 |
17 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
18 | --nproc_per_node=1 \
19 | --master_port 12348 \
20 | main.py \
21 | --phase test_flops \
22 | --configs ${CONFIGS} --drop_last y \
23 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
24 | --val_batch_size 1 \
25 | --shape 1024 1024 \
26 | --test_dir ${DATA_DIR}/val/image \
27 | --data_dir ${DATA_DIR}
28 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrnet_w48_ocr_512x512.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | PYTHON="python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="hrnet48"
12 | MODEL_NAME="hrnet_w48_ocr"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 512 512 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrvit_w18_ocr_v2_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | # PYTHON="python"
7 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
8 |
9 | # check the enviroment info
10 | export PYTHONPATH="$PWD":$PYTHONPATH
11 | DATA_DIR="${DATA_ROOT}/cityscapes"
12 | BACKBONE="hrt18_win11"
13 | MODEL_NAME="hrt_w18_ocr_v2"
14 | CONFIGS="configs/cityscapes/H_48_D_4.json"
15 |
16 |
17 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
18 | --nproc_per_node=1 \
19 | --master_port 12348 \
20 | main.py \
21 | --phase test_flops \
22 | --configs ${CONFIGS} --drop_last y \
23 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
24 | --val_batch_size 1 \
25 | --shape 1024 1024 \
26 | --test_dir ${DATA_DIR}/val/image \
27 | --data_dir ${DATA_DIR}
28 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrvit_w32_ocr_v2_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 | PYTHON="python"
6 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="hrt32_win11"
12 | MODEL_NAME="hrt_w32_ocr_v2"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 1024 1024 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
28 | # win9: [422.8918078339999], time: [0.18235397338867188], params: [13538496]
29 | # win11: [426.503305604], time: [0.1842280626296997], params: [13557040]
30 |
31 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrvit_w32_ocr_v2_512x512.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | PYTHON="python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="hrt32"
12 | MODEL_NAME="hrt_w32_ocr_v2"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 512 512 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrvit_w78_ocr_v2_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | # PYTHON="python"
7 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
8 |
9 | # check the enviroment info
10 | export PYTHONPATH="$PWD":$PYTHONPATH
11 | DATA_DIR="${DATA_ROOT}/cityscapes"
12 | BACKBONE="hrt78_win5"
13 | MODEL_NAME="hrt_w78_ocr_v2"
14 | CONFIGS="configs/cityscapes/H_48_D_4.json"
15 |
16 |
17 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
18 | --nproc_per_node=1 \
19 | --master_port 12348 \
20 | main.py \
21 | --phase test_flops \
22 | --configs ${CONFIGS} --drop_last y \
23 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
24 | --val_batch_size 1 \
25 | --shape 1024 1024 \
26 | --test_dir ${DATA_DIR}/val/image \
27 | --data_dir ${DATA_DIR}
28 |
29 | # win5: [1040.3456280679998], time: [0.4938585996627808], params: [56011040]
30 | # win9: [1064.383171876], time: [0.2658221960067749], params: [56061792]
31 | # win11: [1069.619142052], time: [0.26983258724212644], params: [56098880]
32 | # win15: [1119.8634440679998], time: [0.2960237741470337], params: [56196480]
33 | # msw_hrt78_v1: [1063.5940612839997], time: [0.29960753917694094], params: [56116320]
34 |
35 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_hrvit_w78_ocr_v2_512x512.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | # PYTHON="python"
7 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
8 |
9 | # check the enviroment info
10 | export PYTHONPATH="$PWD":$PYTHONPATH
11 | DATA_DIR="${DATA_ROOT}/cityscapes"
12 | BACKBONE="hrt78"
13 | MODEL_NAME="hrt_w78_ocr_v2"
14 | CONFIGS="configs/cityscapes/H_48_D_4.json"
15 |
16 |
17 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
18 | --nproc_per_node=1 \
19 | --master_port 12348 \
20 | main.py \
21 | --phase test_flops \
22 | --configs ${CONFIGS} --drop_last y \
23 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
24 | --val_batch_size 1 \
25 | --shape 512 512 \
26 | --test_dir ${DATA_DIR}/val/image \
27 | --data_dir ${DATA_DIR}
28 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_wide_hrnet_w32_ocr_v2_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | PYTHON="python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="wide_hrnet32"
12 | MODEL_NAME="wide_hrnet_w32_ocr_v2"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 1024 1024 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_wide_hrnet_w32_ocr_v2_512x512.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | PYTHON="python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="wide_hrnet32"
12 | MODEL_NAME="wide_hrnet_w32_ocr_v2"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 512 512 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_wide_hrnet_w78_ocr_v2_1024x1024.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | # PYTHON="python"
7 | PYTHON="/data/anaconda/envs/pytorch1.7.1/bin/python"
8 |
9 | # check the enviroment info
10 | export PYTHONPATH="$PWD":$PYTHONPATH
11 | DATA_DIR="${DATA_ROOT}/cityscapes"
12 | BACKBONE="wide_hrnet78"
13 | MODEL_NAME="wide_hrnet_w78_ocr_v2"
14 | CONFIGS="configs/cityscapes/H_48_D_4.json"
15 |
16 |
17 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
18 | --nproc_per_node=1 \
19 | --master_port 12348 \
20 | main.py \
21 | --phase test_flops \
22 | --configs ${CONFIGS} --drop_last y \
23 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
24 | --val_batch_size 1 \
25 | --shape 1024 1024 \
26 | --test_dir ${DATA_DIR}/val/image \
27 | --data_dir ${DATA_DIR}
28 |
--------------------------------------------------------------------------------
/seg/scripts/test_flops/test_wide_hrnet_w78_ocr_v2_512x512.sh:
--------------------------------------------------------------------------------
1 | #!/usr/bin/env bash
2 |
3 | . config.profile
4 |
5 |
6 | PYTHON="python"
7 |
8 | # check the enviroment info
9 | export PYTHONPATH="$PWD":$PYTHONPATH
10 | DATA_DIR="${DATA_ROOT}/cityscapes"
11 | BACKBONE="wide_hrnet78"
12 | MODEL_NAME="wide_hrnet_w78_ocr_v2"
13 | CONFIGS="configs/cityscapes/H_48_D_4.json"
14 |
15 |
16 | CUDA_VISIBLE_DEVICES=7 $PYTHON -m torch.distributed.launch \
17 | --nproc_per_node=1 \
18 | --master_port 12348 \
19 | main.py \
20 | --phase test_flops \
21 | --configs ${CONFIGS} --drop_last y \
22 | --backbone ${BACKBONE} --model_name ${MODEL_NAME} \
23 | --val_batch_size 1 \
24 | --shape 512 512 \
25 | --test_dir ${DATA_DIR}/val/image \
26 | --data_dir ${DATA_DIR}
27 |
--------------------------------------------------------------------------------
/seg/segmentor/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/segmentor/__init__.py
--------------------------------------------------------------------------------
/seg/segmentor/tools/__init__.py:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/HRNet/HRFormer/1245b88b5824fbd8cdb358b5ee909a4e537a2ef5/seg/segmentor/tools/__init__.py
--------------------------------------------------------------------------------
/seg/segmentor/tools/evaluator/__init__.py:
--------------------------------------------------------------------------------
1 | import os
2 |
3 | from lib.utils.tools.logger import Logger as Log
4 | from . import standard
5 |
6 | evaluators = {
7 | 'standard': standard.StandardEvaluator
8 | }
9 |
10 |
11 | def get_evaluator(configer, trainer, name=None):
12 | name = os.environ.get('evaluator', 'standard')
13 |
14 | if not name in evaluators:
15 | raise RuntimeError('Unknown evaluator name: {}'.format(name))
16 | klass = evaluators[name]
17 | Log.info('Using evaluator: {}'.format(klass.__name__))
18 |
19 | return klass(configer, trainer)
20 |
--------------------------------------------------------------------------------
|