============= This code was used for experiments with ID-discriminative Embedding (IDE) for Market-1501 dataset.
Thanks Liboyue, give us suggestions for improvement.
If you find this code useful in your research, please consider citing:
@article{zheng2016person,
title={Person Re-identification: Past, Present and Future},
author={Zheng, Liang and Yang, Yi and Hauptmann, Alexander G},
journal={arXiv preprint arXiv:1610.02984},
year={2016}
}
@inproceedings{zheng2015scalable,
title={Scalable Person Re-identification: A Benchmark},
author={Zheng, Liang and Shen, Liyue and Tian, Lu and Wang, Shengjin and Wang, Jingdong and Tian, Qi},
booktitle={Computer Vision, IEEE International Conference on},
year={2015}
}
Requirements for Caffe and matcaffe (see: Caffe installation instructions)
- Clone the IDE repository
# Make sure to clone with --recursive
git clone --recursive https://github.com/zhunzhong07/IDE-baseline-Market-1501-
Build Caffe and matcaffe
cd $IDE_ROOT/caffe # Now follow the Caffe installation instructions here: # http://caffe.berkeleyvision.org/installation.html make -j8 && make matcaffe
-
Download pre-computed models and Market-1501 dataset
Please download the pre-trained imagenet models and put it in the "data/imagenet_models" folder.
Please download Market-1501 dataset and unzip it in the "market_evaluation/dataset" folder. - Training
cd $IDE_ROOT
# train IDE on CaffeNet
./experiments/market/train_IDE_CaffeNet.sh
# train IDE ResNet_50
./experiments/market/train_IDE_ResNet_50.sh
# The IDE models are saved under: "out/market_train"
# If you encounter this problem: bash: ./experiments/market/train_IDE_CaffeNet.sh: Permission denied
# Please execute: chmod 777 -R experiments/- Feature Extraction
cd $IDE_ROOT/market_evaluation
Run Matlab: extract_feature.m
# The IDE features are saved under: "market_evaluation/feat"- Evaluation
Run Matlab: baseline_evaluation_IDE.mYou can download our pre-trained IDE models and IDE features, and put them in the "out_put/market_train" and "market_evaluation/feat" folder, respectively.
Using the models and features above, you can reproduce the results as follows:
| Methods | Rank@1 | mAP |
|---|---|---|
| IDE_CaffeNet + Euclidean | 59.53% | 32.85% |
| IDE_CaffeNet + XQDA | 62.00% | 37.55% |
| IDE_CaffeNet + KISSME | 61.02% | 36.72% |
| IDE_ResNet_50 + Euclidean | 75.62% | 50.68% |
| IDE_ResNet_50 + XQDA | 76.01% | 52.98% |
| IDE_ResNet_50 + KISSME | 77.52% | 53.88% |
If you add a dropout = 0.5 layer after pool5, you will get a better performance for ResNet_50:
| Methods | Rank@1 | mAP |
|---|---|---|
| IDE_ResNet_50 + dropout(0.5) + Euclidean | 78.92% | 55.03% |
| IDE_ResNet_50 + dropout(0.5) + XQDA | 77.35% | 56.01% |
| IDE_ResNet_50 + dropout(0.5) + KISSME | 78.80% | 56.13% |
If you have any questions about this code, please do not hesitate to contact us.
Furthermore, you may check the following codes.