Resolution-invariant Person Re-identification accepted by IJCAI2019 https://arxiv.org/abs/1906.09748
This code is mainly encouraged by https://github.com/KaiyangZhou/deep-person-reid
To accelerate evaluation (10x faster), you can use cython-based evaluation code (developed by luzai). First cd to torchreid/eval_lib, then do make or python setup.py build_ext -i. After that, run python test_cython_eval.py to test if the package is successfully installed.
dataset:
VR dataset, which can be constructed by downsampling the original dataset. If you want to download the dataset please contact snmao@pku.edu.cn . And put the dataset in folder data/ For example, data/vr_market1501/query/XXX
train:
python RIPR.py -d market1501 --optim adam --lr 0.0003 --max-epoch 60 --stepsize 20 40 --train-batch 32 --test-batch 100 --save-dir log/RIPR_train --gpu-devices 0
test:
python RIPR.py --evaluate -d market1501 --test-batch 100 --save-dir log/RIPR_train --gpu-devices 0 --testmodel log/RIPR_train/best_model.pth.tar
model:
the model file can be downloaded from https://pan.baidu.com/s/1aOBPaiFnlS7BRJWnXvaOew , whose password is yov0