├── .gitignore ├── .gitmodules ├── README.md ├── data_analysis ├── bdd100k_dataset_distribution_analysis.ipynb ├── feature_extraction.py ├── kitti_dataset_distribution_analysis.ipynb ├── launcher.sh ├── nuscenes_dataset_distribution_analysis.ipynb ├── plot_functions.ipynb ├── plots │ ├── bdd100k_dataset_likelihood.png │ ├── bdd100k_training_clusters.gif │ ├── bdd100k_training_clusters.png │ ├── dequity_loss.png │ ├── kitti_dataset_likelihood.png │ ├── kitti_training_clusters.gif │ ├── kitti_training_clusters.png │ ├── nuscenes_dataset_likelihood.png │ ├── nuscenes_training_clusters.gif │ ├── nuscenes_training_clusters.png │ ├── tusimple_dataset_likelihood.png │ ├── tusimple_training_clusters.gif │ ├── tusimple_training_clusters.png │ ├── waymo_dataset_likelihood.png │ ├── waymo_training_clusters.gif │ └── waymo_training_clusters.png ├── tusimple_dataset_distribution_analysis.ipynb ├── visualize_kitti3d_pred.ipynb └── waymo_dataset_distribution_analysis.ipynb ├── docs ├── bevformer.md ├── bevfusion.md ├── dd3d.md ├── fig │ └── teaser.png └── getting_started.md └── utils └── kitti_object_dataset_downloader.sh /.gitignore: -------------------------------------------------------------------------------- 1 | # pickle files 2 | *.pkl 3 | 4 | # cache files 5 | *__pycache__ 6 | *.ipynb_checkpoints 7 | -------------------------------------------------------------------------------- /.gitmodules: -------------------------------------------------------------------------------- 1 | [submodule "BEVFormer"] 2 | path = BEVFormer 3 | url = git@github.com:towardsautonomy/BEVFormer.git 4 | [submodule "dd3d"] 5 | path = dd3d 6 | url = git@github.com:towardsautonomy/dd3d.git 7 | [submodule "BevFusion"] 8 | path = BevFusion 9 | url = git@github.com:alchemz/bevfusion.git 10 | 11 | [submodule "bevfusion"] 12 | path = bevfusion 13 | url = https://github.com/alchemz/bevfusion.git 14 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # DatasetEquity: Are All Samples Created Equal? In The Quest For Equity Within Datasets 2 | 3 | *This is the official implementation of the ICCV 2023 Workshop paper: [DatasetEquity: Are All Samples Created Equal? In The Quest For Equity Within Datasets](https://arxiv.org/abs/2308.09878).* 4 | 5 | This paper presents a novel method for addressing data imbalance in machine learning. The method computes sample likelihoods based on image appearance using deep perceptual embeddings and clustering. It then uses these likelihoods to weigh samples differently during training with a proposed **Generalized Focal Loss** function. This loss can be easily integrated with deep learning algorithms. Experiments validate the method's effectiveness across autonomous driving vision datasets including KITTI and nuScenes. The loss function improves state-of-the-art 3D object detection methods, achieving over 200% AP gains on under-represented classes (Cyclist) in the KITTI dataset. The results demonstrate the method is generalizable, complements existing techniques, and is particularly beneficial for smaller datasets and rare classes. 6 | 7 | ![Teaser](docs/fig/teaser.png) 8 | 9 | ## [Getting Started](docs/getting_started.md) 10 | 11 | ## TL;DR 12 | 13 | The concept of this paper is simple: (1) Quantify the likelihood of occurrence for each sample in the training dataset (2) Compute **Generalized Focal Loss** based on the likelihoods (3) Train the model with the new weighted loss function. 14 | 15 | **Generalized Focal Loss** requires computing a loss weight for each sample, called *Dequity Weight*, and can be computed as follows: 16 | 17 | ```python 18 | def dequity_loss_weight(self, p: float, 19 | eta: float=1.0, 20 | gamma: float=5.0 21 | ) -> float: 22 | """Calculate the Dquity Weight. 23 | Args: 24 | p (float): The probability of the sample. 25 | eta (float): The parameter to control the weight. 26 | gamma (float): The parameter to control the weight. 27 | Returns: 28 | float: The dequity loss weight. 29 | """ 30 | return (eta + (1 - p) ** gamma) / (eta + 1) 31 | ``` 32 | 33 | The pseudo-code for the training algorithm is as follows: 34 | 35 | ```python 36 | for sample in dataloader: 37 | # retrieve the sample likelihood 38 | p = get_sample_likelihood(sample) 39 | # compute the DEquity weight 40 | w = dequity_loss_weight(p) 41 | # forward pass 42 | y_hat = model(sample) 43 | # compute the loss 44 | loss = loss_fn(y_hat, sample) 45 | # compute the weighted loss 46 | weighted_loss = w * loss <-- Generalized Focal Loss (Our Contribution) 47 | # backward pass 48 | weighted_loss.backward() 49 | ``` 50 | 51 | Sample likelihoods are computed beforehand. Please refer to the [Getting Started](docs/getting_started.md) guide for more details. 52 | 53 | ## Citation 54 | 55 | If you find this work useful for your research, please cite our paper: 56 | 57 | ``` 58 | @inproceedings{shrivastava2023datasetequity, 59 | title={DatasetEquity: Are All Samples Created Equal? In The Quest For Equity Within Datasets}, 60 | author={Shrivastava, Shubham and Zhang, Xianling and Nagesh, Sushruth and Parchami, Armin}, 61 | booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, 62 | year={2023} 63 | } 64 | ``` 65 | -------------------------------------------------------------------------------- /data_analysis/feature_extraction.py: -------------------------------------------------------------------------------- 1 | # import clustering method 2 | from sklearn.manifold import TSNE 3 | from sklearn.cluster import DBSCAN 4 | from sklearn import metrics 5 | from tqdm.notebook import tqdm 6 | from PIL import Image 7 | 8 | # import torch, cv2 and other dependencies 9 | from torchvision import datasets, transforms 10 | import torch 11 | import torch.nn as nn 12 | from torchvision import transforms 13 | from torchvision.models import resnet101 14 | 15 | import os 16 | import cv2 17 | import pickle 18 | import logging 19 | import numpy as np 20 | import pathlib 21 | import matplotlib.pyplot as plt 22 | import math 23 | import time 24 | import argparse 25 | 26 | # set random seed 27 | import random 28 | random.seed(33) 29 | np.random.seed(33) 30 | 31 | class BDD100kDataset(torch.utils.data.Dataset): 32 | def __init__( 33 | self, 34 | data_root, 35 | image_set="train", 36 | transform=transforms.Compose([transforms.Resize((384, 384)), 37 | transforms.ToTensor(), 38 | transforms.Normalize(0.5, 0.5), 39 | ]), 40 | is_test=False, 41 | keep_difficult=False, 42 | label_file=None, 43 | version='100k', 44 | split='train' 45 | ): 46 | """Dataset for BDD100k data. 47 | Args: 48 | data_root: the root of the BDD100k dataset, 49 | """ 50 | self.data_root = data_root 51 | self.transform = transform 52 | self.version = version 53 | self.split = split 54 | self.filenames, self.tokens = [], [] 55 | 56 | if self.version in ['10k', '100k']: 57 | self.data_root = os.path.join(self.data_root, self.version, self.split) 58 | else: 59 | assert self.version in ['10k','100k'] 60 | print('Please use either 10k samples version or 100k full dataset') 61 | 62 | #pbar = tqdm(enumerate(os.listdir(self.data_root))) 63 | #for filename in os.listdir(self.data_dir): 64 | img_list = os.listdir(self.data_root) 65 | for idx, filename in enumerate(tqdm(img_list)): 66 | img_path = os.path.join(self.data_root, filename) 67 | self.filenames.append(img_path) 68 | self.tokens.append(idx) 69 | 70 | print('Number of data samples in the set: {}'.format(len(self.filenames))) 71 | 72 | # method to get length of data 73 | def __len__(self): 74 | return len(self.filenames) 75 | 76 | # method to get a sample 77 | def __getitem__(self, idx): 78 | # get image 79 | image_filename = self.filenames[idx] 80 | image = Image.open(image_filename).convert('RGB') 81 | # transform image 82 | image = self.transform(image) 83 | # get token 84 | token = self.tokens[idx] 85 | # return image 86 | return {'image': image, 'token': token} 87 | 88 | 89 | 90 | if __name__ == "__main__": 91 | parser = argparse.ArgumentParser() 92 | parser.add_argument('--data_root', type=str, help="path to load the data") 93 | parser.add_argument('--results_dir', type=str, help="folder to save the embeddings") 94 | parser.add_argument('--feature_extactor', type=str, help="model to extract the embeddings") 95 | parser.add_argument('--data_type', type=str, help="select 10k or 100k") 96 | 97 | 98 | args = parser.parse_args() 99 | 100 | results_dir_timestamp = os.path.join(args.results_dir, time.strftime("%Y%m%d_%H%M%S")) 101 | if not os.path.exists(results_dir_timestamp): 102 | os.makedirs(results_dir_timestamp) 103 | print('INFO: {} created to store retrieval results \n'.format(results_dir_timestamp)) 104 | 105 | if args.data_type == 'bdd100k': 106 | dataset = BDD100kDataset(data_root=args.data_root, version='100k', split='train') 107 | else: 108 | pass 109 | 110 | 111 | dataset_str='bdd100k' 112 | train_features_fname = os.path.join(args.results_dir, f'{dataset_str}_training_features.pkl') 113 | train_cluster_info_fname = os.path.join(args.results_dir, f'{dataset_str}_training_cluster_info.pkl') 114 | 115 | 116 | if args.feature_extactor == 'resnet101': 117 | # use a pre-trained model to get the feature vectors for all the images 118 | # in the dataset 119 | model = resnet101(pretrained=True, progress=True) 120 | # remove the last layer keeping the weights 121 | model = torch.nn.Sequential(*list(model.children())[:-1]) 122 | 123 | model.eval() 124 | model.cuda() 125 | train_tokens, train_features = [], [] 126 | with torch.no_grad(): 127 | for i in tqdm(range(len(dataset))): 128 | img = dataset[i]['image'].unsqueeze(0).cuda() 129 | # inference 130 | feature = model(img).flatten().cpu().numpy() 131 | train_features.append(feature) 132 | train_tokens.append(dataset[i]['token']) 133 | train_features = np.array(train_features) 134 | # save the features to disk as a pickle file 135 | with open(train_features_fname, 'wb') as f: 136 | pickle.dump({'tokens': train_tokens, 'features': train_features}, f) 137 | else: 138 | pass 139 | 140 | 141 | -------------------------------------------------------------------------------- /data_analysis/launcher.sh: -------------------------------------------------------------------------------- 1 | export BDD100K_DIR=/s/dat/UserFolders/xzhan258/LaneDetection/BDD100K/bdd100k_images/bdd100k/images 2 | export OUT_DIR=/s/dat/UserFolders/xzhan258/private_dev/DatasetEquity/data_analysis/bdd100k 3 | export ROOT=/s/dat/UserFolders/xzhan258/private_dev/DatasetEquity/data_analysis 4 | 5 | python $ROOT/feature_extraction.py \ 6 | --data_root $BDD100K_DIR \ 7 | --results_dir $OUT_DIR \ 8 | --feature_extactor resnet101 \ 9 | --data_type bdd100k 10 | 11 | 12 | # cd /s/dat/UserFolders/xzhan258/private_dev/DatasetEquity/data_analysis && runpytorch -J feature_extraction -NGPUS 1 -i harbor.hpc.ford.com/xzhan258/torch:1.10_cuda11.4_yolop -x launcher.sh 13 | -------------------------------------------------------------------------------- /data_analysis/plot_functions.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": 1, 6 | "id": "45e31442-613b-4731-85f4-f8fae202e6fa", 7 | "metadata": {}, 8 | "outputs": [], 9 | "source": [ 10 | "# import packages\n", 11 | "import numpy as np\n", 12 | "import matplotlib.pyplot as plt" 13 | ] 14 | }, 15 | { 16 | "cell_type": "code", 17 | "execution_count": 2, 18 | "id": "473eb0c2-9e89-480e-8522-fb1eb3c76343", 19 | "metadata": {}, 20 | "outputs": [], 21 | "source": [ 22 | "def dequity_loss(eta: float, p: float, gamma: float) -> float:\n", 23 | " return (eta + (1 - p) ** gamma) / (eta + 1)" 24 | ] 25 | }, 26 | { 27 | "cell_type": "code", 28 | "execution_count": 4, 29 | "id": "9f849b29-7707-4be1-bcce-c187ab881c00", 30 | "metadata": {}, 31 | "outputs": [ 32 | { 33 | "data": { 34 | "image/png": "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\n", 35 | "text/plain": [ 36 | "
" 37 | ] 38 | }, 39 | "metadata": { 40 | "needs_background": "light" 41 | }, 42 | "output_type": "display_data" 43 | } 44 | ], 45 | "source": [ 46 | "# plot functions for various values of eta and gamma\n", 47 | "fig, ax = plt.subplots(figsize=(10, 8))\n", 48 | "ax.set_xlim(0, 1)\n", 49 | "ax.set_ylim(0, 1)\n", 50 | "ax.set_xlabel('Sample Likelihood', fontsize=16)\n", 51 | "ax.set_ylabel('Loss Weighing Factor', fontsize=16)\n", 52 | "ax.set_title('Generalized Focal Loss', fontsize=20)\n", 53 | "ax.grid(alpha=0.5, linestyle='--', linewidth=1)\n", 54 | "\n", 55 | "eta = 0.0\n", 56 | "p = np.linspace(0, 1, 100)\n", 57 | "gamma = 2\n", 58 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 0.0, $\\gamma$ = 2', linewidth=4)\n", 59 | "\n", 60 | "eta = 0.0\n", 61 | "p = np.linspace(0, 1, 100)\n", 62 | "gamma = 5\n", 63 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 0.0, $\\gamma$ = 5', linewidth=4)\n", 64 | "\n", 65 | "eta = 0.3\n", 66 | "p = np.linspace(0, 1, 100)\n", 67 | "gamma = 2\n", 68 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 0.3, $\\gamma$ = 2', linewidth=4)\n", 69 | "\n", 70 | "eta = 0.3\n", 71 | "p = np.linspace(0, 1, 100)\n", 72 | "gamma = 5\n", 73 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 0.3, $\\gamma$ = 5', linewidth=4)\n", 74 | "\n", 75 | "eta = 1.0\n", 76 | "p = np.linspace(0, 1, 100)\n", 77 | "gamma = 2\n", 78 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 1.0 $\\gamma$ = 2', linewidth=4)\n", 79 | "\n", 80 | "eta = 1.0\n", 81 | "p = np.linspace(0, 1, 100)\n", 82 | "gamma = 5\n", 83 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 1.0, $\\gamma$ = 5', linewidth=4)\n", 84 | "\n", 85 | "eta = 3\n", 86 | "p = np.linspace(0, 1, 100)\n", 87 | "gamma = 2\n", 88 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 3.0, $\\gamma$ = 2', linewidth=4)\n", 89 | "\n", 90 | "eta = 3\n", 91 | "p = np.linspace(0, 1, 100)\n", 92 | "gamma = 5\n", 93 | "ax.plot(p, dequity_loss(eta, p, gamma), label=r'$\\eta$ = 3.0, $\\gamma$ = 5', linewidth=4)\n", 94 | "\n", 95 | "# eta = 0\n", 96 | "# p = np.linspace(0, 1, 100)\n", 97 | "# gamma = 1\n", 98 | "# ax.plot(p, dequity_loss(eta, p, gamma), label='eta = 0.0, gamma = 1')\n", 99 | "\n", 100 | "# fancy legend\n", 101 | "plt.legend(loc='upper right', fancybox=True, framealpha=0.8, fontsize=14, \n", 102 | " ncol=2, facecolor='white', edgecolor='black', shadow=True)\n", 103 | "# shadow\n", 104 | "ax.spines['top'].set_linewidth(0.5)\n", 105 | "ax.spines['right'].set_linewidth(0.5)\n", 106 | "ax.spines['bottom'].set_linewidth(1.5)\n", 107 | "ax.spines['left'].set_linewidth(1.5)\n", 108 | "plt.savefig('plots/dequity_loss.png', dpi=300, bbox_inches='tight')\n", 109 | "plt.show()" 110 | ] 111 | }, 112 | { 113 | "cell_type": "code", 114 | "execution_count": null, 115 | "id": "103cf83e-2f33-4860-a6a7-54d1e70fa112", 116 | "metadata": {}, 117 | "outputs": [], 118 | "source": [] 119 | } 120 | ], 121 | "metadata": { 122 | "kernelspec": { 123 | "display_name": "cdt3d", 124 | "language": "python", 125 | "name": "cdt3d" 126 | }, 127 | "language_info": { 128 | "codemirror_mode": { 129 | "name": "ipython", 130 | "version": 3 131 | }, 132 | "file_extension": ".py", 133 | "mimetype": "text/x-python", 134 | "name": "python", 135 | "nbconvert_exporter": "python", 136 | "pygments_lexer": "ipython3", 137 | "version": "3.7.0" 138 | } 139 | }, 140 | "nbformat": 4, 141 | "nbformat_minor": 5 142 | } 143 | -------------------------------------------------------------------------------- /data_analysis/plots/bdd100k_dataset_likelihood.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/bdd100k_dataset_likelihood.png -------------------------------------------------------------------------------- /data_analysis/plots/bdd100k_training_clusters.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/bdd100k_training_clusters.gif -------------------------------------------------------------------------------- /data_analysis/plots/bdd100k_training_clusters.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/bdd100k_training_clusters.png -------------------------------------------------------------------------------- /data_analysis/plots/dequity_loss.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/dequity_loss.png -------------------------------------------------------------------------------- /data_analysis/plots/kitti_dataset_likelihood.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/kitti_dataset_likelihood.png -------------------------------------------------------------------------------- /data_analysis/plots/kitti_training_clusters.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/kitti_training_clusters.gif -------------------------------------------------------------------------------- /data_analysis/plots/kitti_training_clusters.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/kitti_training_clusters.png -------------------------------------------------------------------------------- /data_analysis/plots/nuscenes_dataset_likelihood.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/nuscenes_dataset_likelihood.png -------------------------------------------------------------------------------- /data_analysis/plots/nuscenes_training_clusters.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/nuscenes_training_clusters.gif -------------------------------------------------------------------------------- /data_analysis/plots/nuscenes_training_clusters.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/nuscenes_training_clusters.png -------------------------------------------------------------------------------- /data_analysis/plots/tusimple_dataset_likelihood.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/tusimple_dataset_likelihood.png -------------------------------------------------------------------------------- /data_analysis/plots/tusimple_training_clusters.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/tusimple_training_clusters.gif -------------------------------------------------------------------------------- /data_analysis/plots/tusimple_training_clusters.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/tusimple_training_clusters.png -------------------------------------------------------------------------------- /data_analysis/plots/waymo_dataset_likelihood.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/waymo_dataset_likelihood.png -------------------------------------------------------------------------------- /data_analysis/plots/waymo_training_clusters.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/waymo_training_clusters.gif -------------------------------------------------------------------------------- /data_analysis/plots/waymo_training_clusters.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/data_analysis/plots/waymo_training_clusters.png -------------------------------------------------------------------------------- /docs/bevformer.md: -------------------------------------------------------------------------------- 1 | ## Training BEVFormer 2 | 3 | Run the docker container in interactive mode. 4 | 5 | ```bash 6 | cd BEVFormer/docker 7 | sh run-docker.sh 8 | ``` 9 | 10 | Prepare the pre-trained model. 11 | 12 | ```bash 13 | cd BEVFormer 14 | mkdir ckpts 15 | 16 | cd ckpts & wget https://github.com/zhiqi-li/storage/releases/download/v1.0/r101_dcn_fcos3d_pretrain.pth 17 | ``` 18 | 19 | Prepare the nuScenes dataset. 20 | 21 | Download `Full dataset (v1.0)` and `CAN bus expansion` from [here](https://www.nuscenes.org/download) and unzip. 22 | 23 | ```bash 24 | mkdir data 25 | cd data 26 | # create symlink to the dataset 27 | ln -s /path/to/nuScenes/data/sets/nuscenes nuscenes 28 | # create symlink to the CAN bus data 29 | ln -s /path/to/nuScenes/data/sets/can_bus can_bus 30 | cd .. 31 | # prepare dataset 32 | python tools/create_data.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag nuscenes --version v1.0 --canbus ./data 33 | # You should see the following structure 34 | BEVFormer 35 | ├── projects/ 36 | ├── tools/ 37 | ├── configs/ 38 | ├── ckpts/ 39 | │ ├── r101_dcn_fcos3d_pretrain.pth 40 | ├── data/ 41 | │ ├── can_bus/ 42 | │ ├── nuscenes/ 43 | │ │ ├── maps/ 44 | │ │ ├── samples/ 45 | │ │ ├── sweeps/ 46 | │ │ ├── v1.0-test/ 47 | | | ├── v1.0-trainval/ 48 | | | ├── nuscenes_infos_temporal_train.pkl 49 | | | ├── nuscenes_infos_temporal_val.pkl 50 | ``` 51 | 52 | Train the baseline model (`variants={base|small|tiny}`). 53 | 54 | ```bash 55 | python tools/train.py projects/configs/bevformer/bevformer_{variant}.py --deterministic 56 | ``` 57 | 58 | If using a launcher for distributed training, use the following command. 59 | 60 | ```bash 61 | python tools/train.py projects/configs/bevformer/bevformer_{variant}.py --deterministic --launcher pytorch 62 | ``` 63 | 64 | For training the model with dataset equity configurations, first change the parameters `model.pts_bbox_head.dequity_eta` and `model.pts_bbox_head.dequity_gamma` in the config file (`projects/configs/bevformer/bevformer_{variant}_de.py`) to the desired values. 65 | 66 | ```bash 67 | python tools/train.py projects/configs/bevformer/bevformer_{variant}_de.py --deterministic --launcher pytorch 68 | ``` -------------------------------------------------------------------------------- /docs/bevfusion.md: -------------------------------------------------------------------------------- 1 | Build the docker using the following command: 2 | 3 | ```bash 4 | cd bevfusion/docker && docker build . -t bevfusion 5 | ``` 6 | 7 | We can then run the docker with the following command: 8 | 9 | ```bash 10 | nvidia-docker run -it -v `pwd`/../data:/dataset --shm-size 16g bevfusion /bin/bash 11 | ``` 12 | Then clone the bevfusion submodule and install it: 13 | 14 | ```bash 15 | cd home && git clone https://github.com/alchemz/bevfusion && cd bevfusion 16 | python setup.py develop 17 | ``` 18 | 19 | You can then create a symbolic link `data` to the `/dataset` directory in the docker. 20 | 21 | ### Data Preparation 22 | 23 | #### nuScenes 24 | 25 | Please follow the instructions from [here](https://github.com/open-mmlab/mmdetection3d/blob/master/docs/en/datasets/nuscenes_det.md) to download and preprocess the nuScenes dataset. Please remember to download both detection dataset and the map extension (for BEV map segmentation). After data preparation, you will be able to see the following directory structure (as is indicated in mmdetection3d): 26 | 27 | ``` 28 | mmdetection3d 29 | ├── mmdet3d 30 | ├── tools 31 | ├── configs 32 | ├── data 33 | │ ├── nuscenes 34 | │ │ ├── maps 35 | │ │ ├── samples 36 | │ │ ├── sweeps 37 | │ │ ├── v1.0-test 38 | | | ├── v1.0-trainval 39 | │ │ ├── nuscenes_database 40 | │ │ ├── nuscenes_infos_train.pkl 41 | │ │ ├── nuscenes_infos_val.pkl 42 | │ │ ├── nuscenes_infos_test.pkl 43 | │ │ ├── nuscenes_dbinfos_train.pkl 44 | ``` 45 | ### Training 46 | 47 | To train the camera-only variant for object detection with cbgs dataset, please run: 48 | 49 | ```bash 50 | torchpack dist-run -np 8 python tools/train.py configs/nuscenes/det/centerhead/lssfpn/camera/256x704/swint/default.yaml --model.encoders.camera.backbone.init_cfg.checkpoint pretrained/swint-nuimages-pretrained.pth 51 | ``` 52 | 53 | To train the camera-only variant for object detection without cbgs dataset, please set ```type``` to ```NoCBGSDataset``` [here](https://github.com/alchemz/bevfusion/blob/d08fd86cccefdacc0f19669cf6c3176a1a51d491/configs/nuscenes/default.yaml#L266). Use the same command as above to train the model. 54 | 55 | Use ```dequity_eta``` and ```dequity_gamma``` parameters in [config](https://github.com/alchemz/bevfusion/blob/main/configs/nuscenes/det/centerhead/default.yaml) to change data equity configurations for performing experiments. 56 | -------------------------------------------------------------------------------- /docs/dd3d.md: -------------------------------------------------------------------------------- 1 | ## Training DD3D 2 | 3 | Prepare KITTI dataset. 4 | 5 | ```bash 6 | mkdir -p data/datasets/KITTI3D 7 | cd data/datasets/KITTI3D 8 | # make symlinks to the KITTI dataset 9 | ln -s /path/to/kitti/training training 10 | ln -s /path/to/kitti/testing testing 11 | # download a standard splits subset of KITTI 12 | curl -s https://tri-ml-public.s3.amazonaws.com/github/dd3d/mv3d_kitti_splits.tar | sudo tar xv -C ./ 13 | cd ../../../.. 14 | ``` 15 | 16 | Run the docker container in interactive mode. 17 | 18 | ```bash 19 | cd dd3d/docker 20 | sh run-docker.sh 21 | ``` 22 | 23 | Train the model with various dequity loss configs. 24 | ```bash 25 | ./train_scripts/train_dd3d_kitti_{dequity_config}.sh 26 | ``` 27 | 28 | Options for `dequity_config` are: 29 | ```bash 30 | baseline 31 | eta0.3_gamma2.0 32 | eta0.5_gamma5.0 33 | ... etc. 34 | ``` -------------------------------------------------------------------------------- /docs/fig/teaser.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/towardsautonomy/DatasetEquity/d6898ad2370a9ea27d6d087c4098620e8e476b59/docs/fig/teaser.png -------------------------------------------------------------------------------- /docs/getting_started.md: -------------------------------------------------------------------------------- 1 | # Getting Started 2 | 3 | ## Installation 4 | 5 | Clone the repository and submodules. 6 | 7 | ```bash 8 | git clone https://github.com/towardsautonomy/DatasetEquity.git --recursive 9 | cd DatasetEquity/BEVFormer 10 | git checkout dataset_equity 11 | cd .. 12 | cd dd3d 13 | git checkout dataset_equity 14 | cd .. 15 | ``` 16 | 17 | Build docker images for `BEVFormer` and `dd3d`. 18 | 19 | ```bash 20 | cd BEVFormer/docker 21 | sh build-docker.sh 22 | cd ../.. 23 | cd dd3d/docker 24 | sh build-docker.sh 25 | cd ../.. 26 | ``` 27 | 28 | **Note**: These docker images are already built and available on [Docker Hub](https://hub.docker.com/u/towardsautonomy). You can skip this step if you don't want to build the docker images yourself. Running the docker images will automatically pull the images from Docker Hub. 29 | 30 | ## Dataset Analysis 31 | 32 | Perform dataset analysis on the KITTI dataset by opening up the [this notebook](/data_analysis/kitti_dataset_distribution_analysis.ipynb), and running the cells. It will generate a file `data_analysis/kitti_training_cluster_info.pkl` which contains all the information needed. By default, this should be copied over to `dd3d/data/datasets/KITTI3D/`. 33 | 34 | For the nuScenes dataset, use [this notebook](/data_analysis/nuscenes_dataset_distribution_analysis.ipynb). It will generate a file `data_analysis/nuscenes_training_cluster_info.pkl`, which should be copied over to `BEVFormer/data/nuscenes/`. These paths can be changed in the project config files. 35 | 36 | 37 | ## [Training BEVFormer](bevformer.md) 38 | 39 | ## [Training DD3D](dd3d.md) 40 | 41 | ## [Training BEVFusion](bevfusion.md) 42 | -------------------------------------------------------------------------------- /utils/kitti_object_dataset_downloader.sh: -------------------------------------------------------------------------------- 1 | #!/bin/sh 2 | 3 | # 4 | # Please note the license conditions of this software / dataset! 5 | # 6 | 7 | # left images 8 | wget https://s3.eu-central-1.amazonaws.com/avg-kitti/data_object_image_2.zip 9 | # right images 10 | wget https://s3.eu-central-1.amazonaws.com/avg-kitti/data_object_image_3.zip 11 | # calibration 12 | wget https://s3.eu-central-1.amazonaws.com/avg-kitti/data_object_calib.zip 13 | # labels 14 | wget https://s3.eu-central-1.amazonaws.com/avg-kitti/data_object_label_2.zip 15 | # unzip all 16 | for z in *.zip; do unzip $z; done --------------------------------------------------------------------------------