- Pytorch 1.1.0
- Torchvision 0.3.0
- Python3.7
- pretrainedmodels
git clone https://github.com/zdaiot/NAIC-Person-Re-identification.git
cd NAIC-Person-Re-identification
Download dataset, unzip and put them into ../Input
directory.
Structure of the ../Input folder can be like:
初赛训练集
初赛A榜测试集
初赛B榜测试集
submission_example_A.json
Create soft links of datasets in the following directories:
cd dataset
mkdir NAIC_data
cd NAIC_data
ln -s ../../../Input/初赛训练集/ ./初赛训练集
ln -s ../../../Input/初赛A榜测试集/ ./初赛A榜测试集
ln -s ../../../Input/submission_example_A.json ./submission_example_A.json
ln -s ~/.cache/torch/checkpoints/resnet50-19c8e357.pth ./
ln -s ../../../Input/初赛B榜测试集/ ./初赛B榜测试集
pip install Cython
cd evaluate/eval_cylib
make
# if you want to test the code
cd ../..
python evaluate/eval_cylib/test_cython.py
python train_baseline.py
python train.py