tonysy / cvpods

All-in-one Toolbox for Computer Vision Research.

Welcome to cvpods, a versatile and efficient codebase for many computer vision tasks: classification, segmentation, detection, self-supervised learning, keypoints and 3D(classification / segmentation / detection / representation learing), etc. The aim of cvpods is to achieve efficient experiments management and smooth tasks-switching.

Each sub-image denotes a task. All images are from search engine.

Table of Contents

• Changelog
• Install
• Usage
  • Get started
  • Step-by-step tutorial
• Model Zoo
• Contributing
• License
• Citation
• Acknowledgement

Changelog

• Dec. 03, 2020: cvpods v0.1 released.

Install

Requirements

• Linux with Python ≥ 3.6
• PyTorch ≥ 1.3 and torchvision that matches the PyTorch installation. You can install them together at pytorch.org to make sure of this
• OpenCV is optional and needed by demo and visualization

Build cvpods from source

Make sure GPU is available on your local machine.

# Install cvpods with GPU directly 
pip install 'git+https://github.com/Megvii-BaseDetection/cvpods.git' --user

# Or, to install it with GPU from a local clone:
git clone https://github.com/Megvii-BaseDetection/cvpods.git
pip install -e cvpods --user 

# Or, to build it without GPU from a local clone:
FORCE_CUDA=1 pip install -e cvpods --user

Usage

Here we demonstrate the basic usage of cvpods (Inference & Train). For more features of cvpods, please refer to our documentation or provided tutorials.

Get Start

Here we use coco object detection task as an example.

# Preprare data path
ln -s /path/to/your/coco/dataset datasets/coco

# Enter a specific experiment dir 
cd playground/retinanet/retinanet.res50.fpn.coco.multiscale.1x

# Train
pods_train --num-gpus 8
# Test
pods_test --num-gpus 8 \
    MODEL.WEIGHTS /path/to/your/save_dir/ckpt.pth # optional
    OUTPUT_DIR /path/to/your/save_dir # optional

# Multi node training
## sudo apt install net-tools ifconfig
pods_train --num-gpus 8 --num-machines N --machine-rank 0/1/.../N-1 --dist-url "tcp://MASTER_IP:port"

Tutorials

We provide a detailed tutorial, which covers introduction, usage, and extend guides in cvpods_tutorials. For all API usages, please refer to our documentation.

Model ZOO

For all the models supported by cvpods, please refer to MODEL_ZOO. We provide 50+ methods across ~15 dataset and ~10 computer vision tasks. cvpods has also supported many research projects of MEGVII Research.

Projects based on cvpods

List is sorted by names.

• AutoAssign
• BorderDet
• DeFCN
• DisAlign
• DynamicHead
• DynamicRouting
• LearnableTreeFilterV2
• LLA
• OTA
• SelfSup
• YOLOF

Contributing

Any kind of contributions (new models / bug report / typo / docs) are welcomed. Please refer to CONTRIBUTING for more details.

License

Apache v2 © Base Detection

Acknowledgement and special thanks

cvpods adopts many components (e.g. network layers) of Detectron2, while cvpods has many advantanges in task support, speed, usability, etc. For more details about official detectron2, please check DETECTRON2

Citing cvpods

If you are using cvpods in your research or wish to refer to the baseline results published in this repo, please use the following BibTeX entry.

@misc{zhu2020cvpods,
  title={cvpods: All-in-one Toolbox for Computer Vision Research},
  author={Zhu*, Benjin and Wang*, Feng and Wang, Jianfeng and Yang, Siwei and Chen, Jianhu and Li, Zeming},
  year={2020}
}