Clear Re-ID: Tackling Occlusion in Person re-identification

This is a project that tries to tackle the problem of occluded images for person re-identification. Our model implements the baseline model from Align Re-ID and modifies the loss and adds a fully connected layer to perform better on occluded images.

Installation

We use Python 2.7 and Pytorch 0.3. For installing Pytorch, follow the official guide. Other packages are specified in requirements.txt.

pip install -r AlignedReID-Re-Production-Pytorc/hrequirements.txt

Dataset

We use Market1501 dataset. Download the dataset from Google Drive. This is a formated dataset derived from the original Market1501 dataset. This can also be created by running the following command on the original Market1501 dataset that can be downloaded from here.

python 	AlignedReID-Re-Production-Pytorch/script/dataset/transform_market1501.py \
--zip_file ~/Dataset/market1501/Market-1501-v15.09.15.zip \
--save_dir ~/Dataset/market1501

Occlusion Simulator

To generate occluded images for the Market1501 dataset, use the following command. It puts the images in /transformed folder inside the source image folder.

python utils/occlusion_simulator.py 
  --d '~/Dataset/market1501/images/' 
  --h 50 
  --w 50 
  --m 'random'
  --partition '/home/ubuntu/Dataset/market1501/partitions.pkl'

Examples

ResNet-50 + Global Loss + OBC Loss on Market1501

Our model works, as of now, only on Market dataset. Use the following command

python AlignedReID-Re-Production-Pytorch/script/experiment/train.py \
-d '(0,)' \
-r 1 \
--dataset market1501 \
--ids_per_batch 16 \
--ims_per_id 4 \
--normalize_feature false \
-gm 0.3 \
-glw 0.8 \
-llw 0 \
-idlw 0 \
-obclw 0.2 \
--base_lr 2e-4 \
--lr_decay_type exp \
--exp_decay_at_epoch 151 \
--total_epochs 300

Log

During training, you can run the TensorBoard and access port 6006 to watch the loss curves etc. E.g.

# Modify the path for `--logdir` accordingly.
tensorboard --logdir YOUR_EXPERIMENT_DIRECTORY/tensorboard

For help regarding TensorBoard.

tensorboard --help

Training Time

Using AWS p2.xlarge instances with 1 GPU took over 15 hours to complete the training.

Testing Time

To test on the Market1501 images (both occluded and non-occluded) it takes ~10 mins on the same AWS instance. For more usage of TensorBoard, see the website and the help:

References & Credits

[1] G. Wang, Y. Yuan, X. Chen, J. Li, and X. Zhou, “Learning discriminative features with multiplegranularities for person re-identification,” 2018.

[2] J. Zhuo, Z. Chen, J. Lai, and G. Wang, “Occluded person re-identification,”arXiv preprintarXiv:1804.02792, 2018.

[3]D. Li, X. Chen, Z. Zhang, and K. Huang, “Learning deep context-aware features over body andlatent parts for person re-identification,” inProceedings of the IEEE Conference on ComputerVision and Pattern Recognition, pp. 384–393, 2017.6

[4]H. Zhao, M. Tian, S. Sun, J. Shao, J. Yan, S. Yi, X. Wang, and X. Tang, “Spindle net: Personre-identification with human body region guided feature decomposition and fusion,”2017 IEEEConference on Computer Vision and Pattern Recognition (CVPR), pp. 907–915, 2017.

[5]S. Bak and P. Carr, “Person re-identification using deformable patch metric learning,”2016 IEEEWinter Conference on Applications of Computer Vision (WACV), pp. 1–9, 2016.

[6]W. Chen, X. Chen, J. Zhang, and K. Huang, “Beyond triplet loss: a deep quadruplet network forperson re-identification,” inThe IEEE Conference on Computer Vision and Pattern Recognition(CVPR), vol. 2, 2017.

[7]X. Zhang, H. Luo, X. Fan, W. Xiang, Y. Sun, Q. Xiao, W. Jiang, C. Zhang, and J. Sun,“Alignedreid: Surpassing human-level performance in person re-identification,”arXiv preprintarXiv:1711.08184, 2017.

[8]A. Hermans, L. Beyer, and B. Leibe, “In defense of the triplet loss for person re-identification,”2017.

[9]L. Zheng, L. Shen, L. Tian, S. Wang, J. Wang, and Q. Tian, “Scalable person re-identification: Abenchmark,”2015 IEEE International Conference on Computer Vision (ICCV), pp. 1116–1124,2015.