EasyCV is an all-in-one computer vision toolbox based on PyTorch, focusing on self-supervised learning, Transformer-based models, and major CV tasks including image classification, metric learning, object detection, pose estimation, and more.

main feature

• SOTA SSL algorithm

  
  

  EasyCV provides state-of-the-art algorithms in self-supervised learning based on contrastive learning, such as SimCLR, MoCO V2, Swav, DINO, and MAE based on mask image modeling. We also provide standard benchmarking tools for ssl model evaluation.

• Vision Transformers

  
  

  EasyCV aims to provide an easy way to use off-the-shelf SOTA transformer models, such as the ViT, Swin Transformer, and DETR series, trained via supervised or self-supervised learning. More models will be added in the future.In addition, we support fromtimmof all pretrained models.

• Functionality and Scalability

  
  

  In addition to SSL, EasyCV also supports image classification, object detection, metric learning, and more fields will be supported in the future. While covering different areas, EasyCV decomposes the framework into different components, such as datasets, models, and run hooks, making it easy to add new components and combine them with existing modules.

  
  

  EasyCV provides a simple and comprehensive inference interface.In addition, PAI-EASAll models are supported, can be easily deployed as an online service, and supports automatic scaling and service monitoring.

• efficiency

  
  

  EasyCV supports multi-GPU and multi-worker training. EasyCV useDALIAccelerate the data io and preprocessing process and useTorchAcceleratorand fp16 to speed up the training process.For inference optimization, EasyCV uses a jit script to export the model, which can be accessed viaPAI-Blade for optimization

what is new

[🔥最新消息] We release our YOLOX-PAI, which can achieve SOTA results in 40~50 mAP (less than 1ms). We also provide a convenient and fast export/predictor API for end-to-end object detection.To get started with YOLOX-PAI quickly, clickhere!

• 31/08/2022 EasyCV v0.6.0 released.

  • Released YOLOX-PAI, achieving SOTA results within 40~50 mAP (less than 1ms)
  • Add detection algorithm DINO, achieve 58.5 mAP on COCO
  • Add mask2former algorithm
  • Use Baidu network disk to release imagenet1k, imagenet22k, coco, lvis, voc2012 data to accelerate download

For more details and history, seechange_log.md.

Install

Please refer to the installationquick_start.mdin the installation section.

start using

Please refer toquick_start.mdGet started quickly. We also provide tutorials for more usage.

notebook

model zoo

Architecture

For more details, please refer to the model zoo below.

data center

EasyCV collects dataset information of different scenarios, which is convenient for users to fine-tune or evaluate models in EasyCV modelzoo.

Please refer todata_hub.md.

license

The project is inApache License (version 2.0)licensed under. The toolkit also contains various third-party components and some code modified from other repos under other open source licenses.For more information, seeNOTICEdocument.

touch

This repo is currently maintained by the PAI-CV team, you can contact us by

EasyCV is an all-in-one computer vision toolbox based on PyTorch, focusing on self-supervised learning, Transformer-based models, and major CV tasks including image classification, metric learning, object detection, pose estimation, and more.

main feature

• SOTA SSL algorithm

  
  

  EasyCV provides state-of-the-art algorithms in self-supervised learning based on contrastive learning, such as SimCLR, MoCO V2, Swav, DINO, and MAE based on mask image modeling. We also provide standard benchmarking tools for ssl model evaluation.

• Vision Transformers

  
  

  EasyCV aims to provide an easy way to use off-the-shelf SOTA transformer models, such as the ViT, Swin Transformer, and DETR series, trained via supervised or self-supervised learning. More models will be added in the future.In addition, we support fromtimmof all pretrained models.

• Functionality and Scalability

  
  

  In addition to SSL, EasyCV also supports image classification, object detection, metric learning, and more fields will be supported in the future. While covering different areas, EasyCV decomposes the framework into different components, such as datasets, models, and run hooks, making it easy to add new components and combine them with existing modules.

  
  

  EasyCV provides a simple and comprehensive inference interface.In addition, PAI-EASAll models are supported, can be easily deployed as an online service, and supports automatic scaling and service monitoring.

• efficiency

  
  

  EasyCV supports multi-GPU and multi-worker training. EasyCV useDALIAccelerate the data io and preprocessing process and useTorchAcceleratorand fp16 to speed up the training process.For inference optimization, EasyCV uses a jit script to export the model, which can be accessed viaPAI-Blade for optimization

what is new

[🔥最新消息] We release our YOLOX-PAI, which can achieve SOTA results in 40~50 mAP (less than 1ms). We also provide a convenient and fast export/predictor API for end-to-end object detection.To get started with YOLOX-PAI quickly, clickhere!

• 31/08/2022 EasyCV v0.6.0 released.

  • Released YOLOX-PAI, achieving SOTA results within 40~50 mAP (less than 1ms)
  • Add detection algorithm DINO, achieve 58.5 mAP on COCO
  • Add mask2former algorithm
  • Use Baidu network disk to release imagenet1k, imagenet22k, coco, lvis, voc2012 data to accelerate download

For more details and history, seechange_log.md.

technical article

We have a series of technical articles about EasyCV features.

Install

Please refer to the installationquick_start.mdin the installation section.

start using

Please refer toquick_start.mdGet started quickly. We also provide tutorials for more usage.

notebook

model zoo

Architecture

For more details, please refer to the model zoo below.

data center

EasyCV collects dataset information of different scenarios, which is convenient for users to fine-tune or evaluate models in EasyCV modelzoo.

Please refer todata_hub.md.

license

The project is inApache License (version 2.0)licensed under. The toolkit also contains various third-party components and some code modified from other repos under other open source licenses.For more information, seeNOTICEdocument.

touch

This repo is currently maintained by the PAI-CV team, you can contact us by

Corporate Services

If you need EasyCV enterprise service support, or purchase cloud product services, you can contact us through DingTalk Group.