diff --git a/README.md b/README.md index c74a68982..5cb38032c 100644 --- a/README.md +++ b/README.md @@ -7,7 +7,8 @@ PaddleClas is a toolset for image classification tasks prepared for the industry and academia. It helps users train better computer vision models and apply them in real scenarios. **Recent update** -- 2021.01.27 Add `ViT` and `DeiT` pretrained model, `ViT`'s Top-1 Acc on ImageNet-1k dataset reaches 85.13%, and `DeiT` reaches 85.1%. +- 2021.02.01 Add `RepVGG` pretrained models, whose Top-1 Acc on ImageNet-1k dataset reaches 79.65%. +- 2021.01.27 Add `ViT` and `DeiT` pretrained models, `ViT`'s Top-1 Acc on ImageNet-1k dataset reaches 85.13%, and `DeiT` reaches 85.1%. - 2021.01.08 Add support for whl package and its usage, Model inference can be done by simply install paddleclas using pip. - 2020.12.16 Add support for TensorRT when using cpp inference to obain more obvious acceleration. - 2020.12.06 Add `SE_HRNet_W64_C_ssld` pretrained model, whose Top-1 Acc on ImageNet-1k dataset reaches 84.75%. @@ -68,6 +69,7 @@ PaddleClas is a toolset for image classification tasks prepared for the industry - [EfficientNet and ResNeXt101_wsl series](#EfficientNet_and_ResNeXt101_wsl_series) - [ResNeSt and RegNet series](#ResNeSt_and_RegNet_series) - [Transformer series](#Transformer) + - [RepVGG series](#RepVGG) - [Others](#Others) - HS-ResNet: arxiv link: [https://arxiv.org/pdf/2010.07621.pdf](https://arxiv.org/pdf/2010.07621.pdf). Code and models are coming soon! - Model training/evaluation @@ -356,7 +358,7 @@ Accuracy and inference time metrics of ResNeSt and RegNet series models are show ### Transformer series -Accuracy and inference time metrics of ViT and DeiT series models are shown as follows. More detailed information can be refered to [Transformer series tutorial](./docs/en/models/Transformer.md). +Accuracy and inference time metrics of ViT and DeiT series models are shown as follows. More detailed information can be refered to [Transformer series tutorial](./docs/en/models/Transformer_en.md). | Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | Flops(G) | Params(M) | Download Address | @@ -384,6 +386,26 @@ Accuracy and inference time metrics of ViT and DeiT series models are shown as f | | | | | | | | | + + +### RepVGG + +Accuracy and inference time metrics of RepVGG series models are shown as follows. More detailed information can be refered to [RepVGG series tutorial](./docs/en/models/RepVGG_en.md). + + +| Model | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | Flops(G) | Params(M) | Download Address | +|------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------| +| RepVGG_A0 | 0.7131 | 0.9016 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams) | +| RepVGG_A1 | 0.7380 | 0.9146 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A1_pretrained.pdparams) | +| RepVGG_A2 | 0.7571 | 0.9264 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A2_pretrained.pdparams) | +| RepVGG_B0 | 0.7450 | 0.9213 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B0_pretrained.pdparams) | +| RepVGG_B1 | 0.7773 | 0.9385 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1_pretrained.pdparams) | +| RepVGG_B2 | 0.7813 | 0.9410 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2_pretrained.pdparams) | +| RepVGG_B1g2 | 0.7732 | 0.9359 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g2_pretrained.pdparams) | +| RepVGG_B1g4 | 0.7675 | 0.9335 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g4_pretrained.pdparams) | +| RepVGG_B2g4 | 0.7881 | 0.9448 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g4_pretrained.pdparams) | +| RepVGG_B3g4 | 0.7965 | 0.9485 | | | | | [Download link](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g4_pretrained.pdparams) | + ### Others @@ -416,4 +438,4 @@ Contributions are highly welcomed and we would really appreciate your feedback!! - Thank [nblib](https://github.com/nblib) to fix bug of RandErasing. - Thank [chenpy228](https://github.com/chenpy228) to fix some typos PaddleClas. -- Thank [jm12138](https://github.com/jm12138) to add ViT and DeiT models into PaddleClas. +- Thank [jm12138](https://github.com/jm12138) to add ViT, DeiT models and RepVGG models into PaddleClas. diff --git a/README_cn.md b/README_cn.md index cb9b81353..bd54999cb 100644 --- a/README_cn.md +++ b/README_cn.md @@ -8,6 +8,7 @@ **近期更新** +- 2021.02.01 添加`RepVGG`系列模型，在ImageNet-1k上Top-1 Acc可达79.65%。 - 2021.01.27 添加`ViT`与`DeiT`模型，在ImageNet-1k上，`ViT`模型Top-1 Acc可达85.13%，`DeiT`模型可达85.1%。 - 2021.01.08 添加whl包及其使用说明，直接安装paddleclas whl包，即可快速完成模型预测。 - 2020.12.16 添加对cpp预测的tensorRT支持，预测加速更明显。 @@ -68,6 +69,7 @@ - [EfficientNet与ResNeXt101_wsl系列](#EfficientNet与ResNeXt101_wsl系列) - [ResNeSt与RegNet系列](#ResNeSt与RegNet系列) - [Transformer系列](#Transformer系列) + - [RepVGG系列](#RepVGG系列) - [其他模型](#其他模型) - HS-ResNet: arxiv文章链接: [https://arxiv.org/pdf/2010.07621.pdf](https://arxiv.org/pdf/2010.07621.pdf)。 代码和预训练模型即将开源，敬请期待。 - 模型训练/评估 @@ -386,6 +388,26 @@ ViT（Vision Transformer）与DeiT（Data-efficient Image Transformers）系列 | | | | | | | | | + + +### RepVGG系列 + +关于RepVGG系列模型的精度、速度指标如下表所示，更多介绍可以参考：[RepVGG系列模型文档](./docs/zh_CN/models/RepVGG.md)。 + + +| 模型 | Top-1 Acc | Top-5 Acc | time(ms)
bs=1 | time(ms)
bs=4 | Flops(G) | Params(M) | 下载地址 | +|------------------------|-----------|-----------|------------------|------------------|----------|-----------|------------------------------------------------------------------------------------------------------| +| RepVGG_A0 | 0.7131 | 0.9016 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A0_pretrained.pdparams) | +| RepVGG_A1 | 0.7380 | 0.9146 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A1_pretrained.pdparams) | +| RepVGG_A2 | 0.7571 | 0.9264 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_A2_pretrained.pdparams) | +| RepVGG_B0 | 0.7450 | 0.9213 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B0_pretrained.pdparams) | +| RepVGG_B1 | 0.7773 | 0.9385 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1_pretrained.pdparams) | +| RepVGG_B2 | 0.7813 | 0.9410 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2_pretrained.pdparams) | +| RepVGG_B1g2 | 0.7732 | 0.9359 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g2_pretrained.pdparams) | +| RepVGG_B1g4 | 0.7675 | 0.9335 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B1g4_pretrained.pdparams) | +| RepVGG_B2g4 | 0.7881 | 0.9448 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B2g4_pretrained.pdparams) | +| RepVGG_B3g4 | 0.7965 | 0.9485 | | | | | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/RepVGG_B3g4_pretrained.pdparams) | + ### 其他模型 @@ -416,6 +438,6 @@ ViT（Vision Transformer）与DeiT（Data-efficient Image Transformers）系列 - 非常感谢[nblib](https://github.com/nblib)修正了PaddleClas中RandErasing的数据增广配置文件。 - 非常感谢[chenpy228](https://github.com/chenpy228)修正了PaddleClas文档中的部分错别字。 -- 非常感谢[jm12138](https://github.com/jm12138)为PaddleClas添加ViT和DeiT模型。 +- 非常感谢[jm12138](https://github.com/jm12138)为PaddleClas添加ViT，DeiT系列模型和RepVGG系列模型。 我们非常欢迎你为PaddleClas贡献代码，也十分感谢你的反馈。 diff --git a/docs/en/models/RepVGG_en.md b/docs/en/models/RepVGG_en.md index 089fccbbb..f2171a8fc 100644 --- a/docs/en/models/RepVGG_en.md +++ b/docs/en/models/RepVGG_en.md @@ -16,12 +16,7 @@ RepVGG (Making VGG-style ConvNets Great Again) series model is a simple but powe | RepVGG_B2 | 0.7813 | 0.9410 | 0.7878 | | | RepVGG_B1g2 | 0.7732 | 0.9359 | 0.7778 | | | RepVGG_B1g4 | 0.7675 | 0.9335 | 0.7758 | | -| RepVGG_B2g4 | 0.7782 | 0.9380 | 0.7850 | | +| RepVGG_B2g4 | 0.7881 | 0.9448 | 0.7938 | | +| RepVGG_B3g4 | 0.7965 | 0.9485 | 0.8021 | | -| Models | Top1 | Top5 | Reference
top1 | FLOPS
(G) | -|:--:|:--:|:--:|:--:|:--:| -| RepVGG_B3_200epochs | 0.7987 | 0.9502 | 0.8052 | | -| RepVGG_B2g4_200epochs | 0.7881 | 0.9448 | 0.7938 | | -| RepVGG_B3g4_200epochs | 0.7965 | 0.9485 | 0.8021 | | - -Params, FLOPs, Inference speed and other information are coming soon. \ No newline at end of file +Params, FLOPs, Inference speed and other information are coming soon. diff --git a/docs/en/models/Transformer.md b/docs/en/models/Transformer_en.md similarity index 100% rename from docs/en/models/Transformer.md rename to docs/en/models/Transformer_en.md diff --git a/docs/zh_CN/models/RepVGG.md b/docs/zh_CN/models/RepVGG.md index 7ae56d8e2..7b2d9b2ec 100644 --- a/docs/zh_CN/models/RepVGG.md +++ b/docs/zh_CN/models/RepVGG.md @@ -17,12 +17,7 @@ RepVGG（Making VGG-style ConvNets Great Again）系列模型是由清华大学( | RepVGG_B2 | 0.7813 | 0.9410 | 0.7878 | | | RepVGG_B1g2 | 0.7732 | 0.9359 | 0.7778 | | | RepVGG_B1g4 | 0.7675 | 0.9335 | 0.7758 | | -| RepVGG_B2g4 | 0.7782 | 0.9380 | 0.7850 | | +| RepVGG_B2g4 | 0.7881 | 0.9448 | 0.7938 | | +| RepVGG_B3g4 | 0.7965 | 0.9485 | 0.8021 | | -| Models | Top1 | Top5 | Reference
top1 | FLOPS
(G) | -|:--:|:--:|:--:|:--:|:--:| -| RepVGG_B3_200epochs | 0.7987 | 0.9502 | 0.8052 | | -| RepVGG_B2g4_200epochs | 0.7881 | 0.9448 | 0.7938 | | -| RepVGG_B3g4_200epochs | 0.7965 | 0.9485 | 0.8021 | | - -关于Params、FLOPs、Inference speed等信息，敬请期待。 \ No newline at end of file +关于Params、FLOPs、Inference speed等信息，敬请期待。