PaddleClas Versions Save

A treasure chest for visual classification and recognition powered by PaddlePaddle

v2.5.2

1 month ago

What's Changed

New Contributors

Full Changelog: https://github.com/PaddlePaddle/PaddleClas/compare/v2.5.1...v2.5.2

v2.5.1

1 year ago
  1. support build index by whl;
  2. fix some bugs.

v2.5.0

1 year ago

1.Release PP-ShiTu V2. 2.Release PP-ShiTu V2 android demo. 3.Release PP-ShiTu feature database management tool.

v2.4.0

1 year ago

1.Release Practical Ultra Light-weight image Classification solutions. PULC models inference within 3ms on CPU devices, with accuracy on par with SwinTransformer. 2.Release 9 PULC models including person attribute, traffic sign recognition, text image orientation classification, etc. 3.Release PP-HGNet classification network, which is suitable for gpu devices 4.Release PP-LCNet v2 classification network, which is suitable for cpu devices. 5.Add CSwinTransformer, PVTv2, MobileViT and VAN. 6.Add BoT ReID models.

v2.3.1

2 years ago

1.Update PP-ShiTu model and add 18MB model series. 2.Upgrade the document completely. 3.Add C++ Inference. 4.Add C++ Pipeline Serving mode. 5.Add a demo for Paddle Lite on Android.

2.3.0

2 years ago

1.Add lite weight models, including detection and feature extraction. 2.Add PP-LCNet backbone model, which is super fast on CPU devices. 3.Support PaddleServing and PaddleSlim. 4.Switch Vector Search module to faiss, due to many compatibility feedback. 5.Add PKSampler, which is more stable on multi-card training. 6.Legendary models now can output middleware result. 7.Add DeepHash module, which can compress float feature to binary feature. 8.SwinTransformer, Twins and Deit achieve same accuracy with the origins training from scratch.

2.2.1

2 years ago

1.Add Swin transformer series model. 2.Support static graph training, support dali and fp16 training. 3.Support build feature gallery with batchsize > 1, support add new feature to existing feature gallery. 4.Fix bugs and update document.

2.2.0

2 years ago
  1. Architecture 1.1. ppcls backbones are now separated into two groups: legendary models and model zoo. 1.2. Legendary models inherit from a new base class TheseusLayer, which allows stop at some point or even change architectures
  2. Metric Learning 2.1. Add a lot of metric learning functions, including gears, which can be inserted into arch , and losses. 2.2. PaddleClas now support classification task and metric learning task using the same trainer. You only need switch different configs.
  3. Vector Search 3.1. Intergrate Mobius vector search algorithm.
  4. Applications 4.1. Add new applications: product recognition, logo recognition, car classification, car ReID and cartoon character recognition. 4.2. Add new image recognition pipeline, which contains detection, feature extraction and vector search.
  5. New models 5.1. add LeViT、Twins、TNT、DLA、HarDNet、RedNet models

v2.1.0

3 years ago
  1. Add RexNet, Mixnet, ViT and DeiT deploy model.
  2. Add new Chinese tutorials for different users.
  3. Update whl package.

v2.0.0

3 years ago

Release Note

Support dynamic graph programming paradigm, adapted to Paddle2.0. Including:

  1. 29 series of classification network structures and training configurations, 134 models' pretrained weights and their evaluation metrics.
  2. SSLD Knowledge Distillation. Based on this SSLD distillation strategy, the top-1 acc of the distilled model is generally increased by more than 3%.
  3. Data augmentation: PaddleClas provides detailed introduction of 8 data augmentation algorithms such as AutoAugment, Cutout, Cutmix, code reproduction and effect evaluation in a unified experimental environment.
  4. Pretrained model with 100,000 categories: Based on ResNet50_vd model, Baidu open sourced the ResNet50_vd pretrained model trained on a 100,000-category dataset. In some practical scenarios, the accuracy based on the pretrained weights can be increased by up to 30%.
  5. A variety of training modes, including multi-machine training, mixed precision training, etc.
  6. A variety of inference and deployment solutions, including TensorRT inference, Paddle-Lite inference, model service deployment, model quantification, Paddle Hub, etc.
  7. Support Linux, Windows, macOS and other systems.
  8. Support training/evaluation on CPU/CPU/XPU.