This is forked from https://github.com/jiyanggao/Video-Person-ReID
All available weights are Here

Train:
python main_video_person_reid.py -d viva --arch resnet50tp --save-dir log/resnet50tp_viva
python main_video_person_reid.py -d mars --arch resnet50tp --save-dir log/resnet50tp_mars
Test:
python main_video_person_reid.py -d viva --arch resnet50tp --save-dir log/resnet50tp_viva --resume --evaluate
python main_video_person_reid.py -d mars --arch resnet50tp --save-dir log/resnet50tp_mars --resume --evaluate
Calculating similarity score:
python main_video_person_reid.py --simi --path test_imgs --resume -d viva --arch resnet50tp --save-dir log/resnet50tp_viva
python main_video_person_reid.py --simi --path test_imgs --resume -d mars --arch resnet50tp --save-dir log/resnet50tp_mars
python main_video_person_reid.py --simi --path ../new_vivadata/384568.track --resume -d mars --arch resnet50tp --save-dir log/resnet50tp_mars
Calculating the best threshold:
python main_video_person_reid.py --roc -d mars --resume --arch resnet50tp --save-dir log/resnet50tp_mars
python main_video_person_reid.py --roc -d viva --resume --arch resnet50tp --save-dir log/resnet50tp_viva

See opt.py and below for more detailed instruction of running code on the command line.
-d mars and --save-dir log/resnet50tp_mars and --arch resnet50tp should be consistent: it means you used mars datasets to train resnet50tp model

Video-Person-ReID

This is the code repository for tech report "Revisiting Temporal Modeling for Video-based Person ReID": https://arxiv.org/abs/1805.02104. If you find this help your research, please cite

@article{gao2018revisiting,
  title={Revisiting Temporal Modeling for Video-based Person ReID},
  author={Gao, Jiyang and Nevatia, Ram},
  journal={arXiv preprint arXiv:1805.02104},
  year={2018}
}

Introduction

This repository contains PyTorch implementations of temporal modeling methods for video-based person reID. It is forked from deep-person-reid.. Based on that, I implement (1) video sampling strategy for training and testing, (2) temporal modeling methods including temporal pooling, temporal attention, RNN and 3D conv. The base loss function and basic training framework remain the same as deep-person-reid.

Motivation

Although previous work proposed many temporal modeling methods and did extensive experiments, but it's still hard for us to have an "apple-to-apple" comparison across these methods. As the image-level feature extractor and loss function are not the same, which have large impact on the final performance. Thus, we want to test the representative methods under an uniform framework.

Dataset

All experiments are done on MARS, as it is the largest dataset available to date for video-based person reID. Please follow deep-person-reid to prepare the data. The instructions are copied here:

1. Create a directory named mars/ under ../datasets/.
2. Download dataset to ../datasets/mars/ from http://www.liangzheng.com.cn/Project/project_mars.html.
3. Extract bbox_train.zip and bbox_test.zip.
4. Download split information from https://github.com/liangzheng06/MARS-evaluation/tree/master/info and put info/ in ../datasets/mars/ (we want to follow the standard split in [8]). The data structure would look like:

mars/
    bbox_test/
    bbox_train/
    info/

5. Use -d mars when running the code.

For vivalab's dataset:

1. Create a directory named viva_dataset/viva_dataset/ under ../datasets/.
2. Download dataset to ../datasets/viva_dataset/viva_dataset/
3. Use -d viva when running the code.

Usage

To train the model, please run

python main_video_person_reid.py --arch=resnet50tp

arch could be resnet50tp (Temporal Pooling), resnet50ta (Temporal Attention), resnet50rnn (RNN), resnet503d (3D conv). For 3D conv, I use the design and implementation from 3D-ResNets-PyTorch, just minor modification is done to fit the network into this person reID system.

I found that learning rate has a significant impact on the final performance. Here are the learning rates I used (may not be the best): 0.0003 for temporal pooling, 0.0003 for temporal attention, 0.0001 for RNN, 0.0001 for 3D conv.

Other detailed settings for different temporal modeling could be found in models/ResNet.py

Similarity score

--simi to enable similarity mode
--path test_imgs: For calculating similarity score of tracklets, set args.path as the path of all tracklets, the path structure should be like ./identity_xx/tracklet_xx/xxx.png
The result will be saved in result.csv

Best threshold

--roc to enable this mode, it will produce the best threshold, TP_TN.png and ROC.png.

Performance of the paper

Model	mAP	CMC-1	CMC-5	CMC-10	CMC-20
image-based	74.1	81.3	92.6	94.8	96.7
pooling	75.8	83.1	92.8	95.3	96.8
attention	76.7	83.3	93.8	96.0	97.4
rnn	73.9	81.6	92.8	94.7	96.3
3d conv	70.5	78.5	90.9	93.9	95.9

Performance of my result on Mars

Model	mAP	CMC-1	CMC-5	CMC-10	CMC-20
pooling	76.3	82.2	93.5	95.7	96.7
attention	75.7	82.2	93.3	95.7	97.1

Performance of my result on Viva_dataset

Model	mAP	CMC-1	CMC-5	CMC-10	CMC-20
pooling	93.2	92.2	97.7	97.2	100.0

Mars

log/resnet50tp_mars Best threshold: 285
log/resnet50tp_mars AUC: 0.30726125418013067

Viva

log/resnet50tp_viva Best threshold: 263
log/resnet50tp_viva AUC: 0.5124618830862768

Prerequisites

• Linux kernel 4.15.0-58-generic
• gcc version 5.4.0
• Ubuntu 16.04.6 LTS
• CUDA Version 9.0.176
• Python 3.6.8
• Pytorch 1.1.0

Python Packages Version (conda 4.7.10)

• cudatoolkit 9.0 h13b8566_0
• matplotlib 3.1.0 py36h5429711_0
• numpy 1.16.4 py36h7e9f1db_0
• numpy-base 1.16.4 py36hde5b4d6_0
• opencv 3.4.2 py36h6fd60c2_1
• openssl 1.1.1c h516909a_0 conda-forge
• pandas 0.25.0 py36he6710b0_0
• pillow 6.1.0 py36h34e0f95_0
• pip 19.1.1 py36_0
• python 3.6.8 h0371630_0
• python-dateutil 2.8.0 py36_0
• pytorch 1.1.0 py3.6_cuda9.0.176_cudnn7.5.1_0 pytorch