Epiphqny / VisTR

[CVPR2021 Oral] End-to-End Video Instance Segmentation with Transformers

Home Page:https://arxiv.org/abs/2011.14503

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

VisTR: End-to-End Video Instance Segmentation with Transformers

This is the official implementation of the VisTR paper:


We provide instructions how to install dependencies via conda. First, clone the repository locally:

git clone https://github.com/Epiphqny/vistr.git

Then, install PyTorch 1.6 and torchvision 0.7:

conda install pytorch==1.6.0 torchvision==0.7.0

Install pycocotools

conda install cython scipy
pip install -U 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
pip install git+https://github.com/youtubevos/cocoapi.git#"egg=pycocotools&subdirectory=PythonAPI"

Compile DCN module(requires GCC>=5.3, cuda>=10.0)

cd models/dcn
python setup.py build_ext --inplace


Download and extract 2019 version of YoutubeVIS train and val images with annotations from CodeLab or YoutubeVIS. We expect the directory structure to be the following:

├── data
│   ├── train
│   ├── val
│   ├── annotations
│   │   ├── instances_train_sub.json
│   │   ├── instances_val_sub.json
├── models

Download the pretrained DETR models Google Drive BaiduYun(passcode:alge) on COCO and save it to the pretrained path.


Training of the model requires at least 32g memory GPU, we performed the experiment on 32g V100 card. (As the training resolution is limited by the GPU memory, if you have a larger memory GPU and want to perform the experiment, please contact with me, thanks very much)

To train baseline VisTR on a single node with 8 gpus for 18 epochs, run:

python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --backbone resnet101/50 --ytvos_path /path/to/ytvos --masks --pretrained_weights /path/to/pretrained_path


python inference.py --masks --model_path /path/to/model_weights --save_path /path/to/results.json


We provide baseline VisTR models, and plan to include more in future. AP is computed on YouTubeVIS dataset by submitting the result json file to the CodeLab system, and inference time is calculated by pure model inference time (without data-loading and post-processing).

name backbone FPS mask AP model result json zip detailed AP
0 VisTR R50 69.9 36.2 vistr_r50.pth vistr_r50.zip

1 VisTR R101 57.7 40.1 vistr_r101.pth vistr_r101.zip


VisTR is released under the Apache 2.0 license. Please see the LICENSE file for more information.


We would like to thank the DETR open-source project for its awesome work, part of the code are modified from its project.


Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follow.

  title={End-to-End Video Instance Segmentation with Transformers},
  author={Wang, Yuqing and Xu, Zhaoliang and Wang, Xinlong and Shen, Chunhua and Cheng, Baoshan and Shen, Hao and Xia, Huaxia},
  booktitle =  {Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR)},
ezoic increase your site revenue


[CVPR2021 Oral] End-to-End Video Instance Segmentation with Transformers


License:Apache License 2.0


Language:Python 65.5%Language:Cuda 30.9%Language:C 3.3%Language:C++ 0.3%