Torchreid¶

Torchreid is a library built on PyTorch for deep-learning person re-identification.

It features:

multi-GPU training
support both image- and video-reid
end-to-end training and evaluation
incredibly easy preparation of reid datasets
multi-dataset training
cross-dataset evaluation
standard protocol used by most research papers
highly extensible (easy to add models, datasets, training methods, etc.)
implementations of state-of-the-art deep reid models
access to pretrained reid models
advanced training techniques
visualization tools (tensorboard, ranks, etc.)

Documentation: https://kaiyangzhou.github.io/deep-person-reid/.

Code: https://github.com/KaiyangZhou/deep-person-reid.

Model zoo: https://kaiyangzhou.github.io/deep-person-reid/MODEL_ZOO.

Installation¶

We recommend using conda to manage the packages.

Clone deep-person-reid to your preferred directory.

$ git clone https://github.com/KaiyangZhou/deep-person-reid.git

Create a conda environment (the default name is torchreid).

$ cd deep-person-reid/
$ conda env create -f environment.yml
$ conda activate torchreid

Do check whether which python and which pip point to the right path.

Install tensorboard.

$ pip install tb-nightly

Install PyTorch and torchvision (select the proper cuda version to suit your machine).

$ conda install pytorch torchvision cudatoolkit=9.0 -c pytorch

Install torchreid.

$ python setup.py develop

Get started: 30 seconds to Torchreid¶

Import torchreid

import torchreid

Load data manager

datamanager = torchreid.data.ImageDataManager(
    root='reid-data',
    sources='market1501',
    height=256,
    width=128,
    batch_size=32,
    transforms=['random_flip', 'random_crop']
)

3 Build model, optimizer and lr_scheduler

model = torchreid.models.build_model(
    name='resnet50',
    num_classes=datamanager.num_train_pids,
    loss='softmax',
    pretrained=True
)

model = model.cuda()

optimizer = torchreid.optim.build_optimizer(
    model,
    optim='adam',
    lr=0.0003
)

scheduler = torchreid.optim.build_lr_scheduler(
    optimizer,
    lr_scheduler='single_step',
    stepsize=20
)

Build engine

engine = torchreid.engine.ImageSoftmaxEngine(
    datamanager,
    model,
    optimizer=optimizer,
    scheduler=scheduler,
    label_smooth=True
)

Run training and test

engine.run(
    save_dir='log/resnet50',
    max_epoch=60,
    eval_freq=10,
    print_freq=10,
    test_only=False
)

A unified interface¶

In “deep-person-reid/scripts/”, we provide a unified interface including a default parser file default_parser.py and the main script main.py. For example, to train an image reid model on Market1501 using softmax, you can do

python main.py \
--root path/to/reid-data \
--app image \
--loss softmax \
--label-smooth \
-s market1501 \
-a resnet50 \
--optim adam \
--lr 0.0003 \
--max-epoch 60 \
--stepsize 20 40 \
--batch-size 32 \
--transforms random_flip random_crop \
--save-dir log/resnet50-market1501-softmax \
--gpu-devices 0

Please refer to default_parser.py and main.py for more details.

Datasets¶

Image-reid datasets¶

Video-reid datasets¶

Models¶

ImageNet classification models¶

Lightweight models¶

ReID-specific models¶

Losses¶

Citation¶

If you find this code useful to your research, please cite the following publication.

@article{zhou2019osnet,
  title={Omni-Scale Feature Learning for Person Re-Identification},
  author={Zhou, Kaiyang and Yang, Yongxin and Cavallaro, Andrea and Xiang, Tao},
  journal={arXiv preprint arXiv:1905.00953},
  year={2019}
}