From a6c9f3763d4aca76ec7961746ae1af7889603563 Mon Sep 17 00:00:00 2001 From: Jason Date: Mon, 24 May 2021 10:36:30 +0800 Subject: [PATCH 1/5] Update model_zoo.md --- docs/en/model_zoo.md | 51 +++++++++----------------------------------- 1 file changed, 10 insertions(+), 41 deletions(-) diff --git a/docs/en/model_zoo.md b/docs/en/model_zoo.md index bd761786a..8d35e1dd9 100644 --- a/docs/en/model_zoo.md +++ b/docs/en/model_zoo.md @@ -49,44 +49,13 @@ Support UNet, HRNet, DeepLab and so on. Test models are from PaddleSeg [release/ |UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/unet) | ## object detection -comming soon - -# Static computational graph -## Image classification -Test models are from PaddleClas [master branch](https://github.com/PaddlePaddle/PaddleClas/tree/master). - -| Models | Source | -|-------|--------| -| ResNet | [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/resnet.py) | -| DenseNet | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/densenet.py) | -| ShuffleNet | [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/shufflenet_v2.py) | -| MobileNet| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/mobilenet_v3.py) | -| VGG16| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/vgg.py) | -| SE_ResNext50| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/se_resnext.py) | -| InceptionV4| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/inception_v4.py) | -| SE_ResNet50_vd| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/se_resnext_vd.py) | -| SqueezeNet1_0| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/squeezenet.py) | -| Xception41| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/xception.py) | - -## object detection -Support SSD,YoloV3. Test models are from PaddleDetection [release/0.4 branch](https://github.com/PaddlePaddle/Paddledetection/tree/release/0.4). -Due to ONNX's limit, Paddle2ONNX is not able to support all the detection models and now only supports one-stage detection model. - -| Models | Source | -|-------|--------| -|SSD_MobileNet|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#ssd) | -|YoloV3_DarkNet53|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#yolo-v3) | -|YoloV3_ResNet34|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#yolo-v3) | -|YoloV3_MobileNet|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#yolo-v3) | -|FasterRCNN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#faster--mask-r-cnn) | -|FasterRCNN_FPN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#faster--mask-r-cnn) | -|FasterRCNN_FPN_DCN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO.md#deformable-convnets-v2) | - -## Image segmentation -Support UNet,HRNet and DeepLab. Test models are from PaddleSeg [release/v0.7.0 branch](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v0.7.0). - -| Models | Source | -|-------|--------| -|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_unet.md) | -|HRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_hrnet.md) | -|DeepLab|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_deeplabv3plus.md) | +Support 8 object detection archtectures. Test models are from PaddleDetection [release/2.1](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.1) +| Models | Source | +| ----------- | ------------------------------------------------------------ | +| YOLO-V3 | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/yolov3/ | +| PPYOLO | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| PPYOLO-Tiny | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| PPYOLO-V2 | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| TTFNet | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ttfnet/ | +| PAFNet | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ttfnet/ | +| SSD | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ssd/ | From a4fc85ba9a1174e5ab327ba98cd9d79c8d7a2d7e Mon Sep 17 00:00:00 2001 From: Jason Date: Tue, 25 May 2021 17:03:16 +0800 Subject: [PATCH 2/5] Update model_zoo.md --- docs/en/model_zoo.md | 41 +++++++++++++++++++---------------------- 1 file changed, 19 insertions(+), 22 deletions(-) diff --git a/docs/en/model_zoo.md b/docs/en/model_zoo.md index 8d35e1dd9..c7d37eaaa 100644 --- a/docs/en/model_zoo.md +++ b/docs/en/model_zoo.md @@ -9,20 +9,18 @@ As dynamic computational model is under develop, We will update more convertabl ## Image classification -Comprehensive coverage of image classification models，now we support the whole series model in PaddlClas [dygraph branch](https://github.com/paddlepaddle/PaddleClas/tree/dygraph). +Comprehensive coverage of image classification models，now we support the whole series model in PaddlClas [release/2.1](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1). |Models | Source | |---|---| -| ResNet series| [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#resnet-and-vd-series)| -| Mobile series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#mobile-series)| -| SEResNeXt and Res2Net series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#seresnext-and-res2net-series)| -| DPN and DenseNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#dpn-and-densenet-series)| -| HRNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#hrnet-series)| -| Inception series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#inception-series)| -| EfficientNet and ResNeXt101_wsl series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#efficientnet-and-resnext101_wsl-series)| -| ResNeSt and RegNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#resnest-and-regnet-series)| -| Transformer Series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#transformer-series)| -| Others |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README.md#others)| +| ResNet series| [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#ResNet_and_Vd_series)| +| Mobile series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#Mobile_series)| +| SEResNeXt and Res2Net series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#SEResNeXt_and_Res2Net_series)| +| DPN and DenseNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#DPN_and_DenseNet_series)| +| HRNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#HRNet_series)| +| Inception series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#Inception_series)| +| EfficientNet and ResNeXt101_wsl series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#EfficientNet_and_ResNeXt101_wsl_series)| +| ResNeSt and RegNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#ResNeSt_and_RegNet_series)| ## OCR @@ -30,23 +28,22 @@ Support CRNN(Text Detection Model), DB(Text Recognition Model) and Text Angle Cl | Models | Source | |-------|--------| -|DB|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc_en/algorithm_overview_en.md#1-text-detection-algorithm) | -|CRNN|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_en/algorithm_overview_en.md#2-text-recognition-algorithm) | -|CLS|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_en/models_list_en.md#3-text-angle-classification-model) | +|Chinese and English ultra-lightweight OCR model (9.4M) |[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR#pp-ocr-20-series-model-listupdate-on-dec-15) | +|Chinese and English general OCR model (143.4M)|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR#pp-ocr-20-series-model-listupdate-on-dec-15) | ## Image segmentation Support UNet, HRNet, DeepLab and so on. Test models are from PaddleSeg [release/v2.0.0-rc branch](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.0.0-rc)。 | Models | Source | |-------|--------| -|BiSeNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.0.0-rc/configs/bisenet) | -|DANet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/danet) | -|DeepLabv3|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/deeplabv3) | -|Deeplabv3P |[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/deeplabv3p) | -|FCN|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/fcn) | -|GCNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/gcnet) | -|OCRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/ocrnet) | -|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/unet) | +|BiSeNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.1/configs/bisenet) | +|DANet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/danet) | +|DeepLabv3|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/deeplabv3) | +|Deeplabv3P |[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/deeplabv3p) | +|FCN|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/fcn) | +|GCNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/gcnet) | +|OCRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/ocrnet) | +|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/unet) | ## object detection Support 8 object detection archtectures. Test models are from PaddleDetection [release/2.1](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.1) From 713194e5a7ef397d739ccb3a21cc216bc7fdbb57 Mon Sep 17 00:00:00 2001 From: Jason Date: Tue, 25 May 2021 17:04:04 +0800 Subject: [PATCH 3/5] Update model_zoo.md --- docs/en/model_zoo.md | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/docs/en/model_zoo.md b/docs/en/model_zoo.md index c7d37eaaa..d74a15179 100644 --- a/docs/en/model_zoo.md +++ b/docs/en/model_zoo.md @@ -1,11 +1,7 @@ -Paddle2ONNX mainly supports three types of models: image classification, object detection and image segmentation. -As PaddlePaddle 2.0 evolves, the realization of serialized operators will be more universal. It is expected that NLP and OCR series of models will be supported. +Paddle2ONNX supports converting PaddlePaddle model to ONNX format. Due to the differences between frameworks, some models may not be supported. If you meet any problem such as converting failure or inference error，you can raise a issue in [ISSUE](https://github.com/PaddlePaddle/paddle-onnx/issues/new). -# Dynamic computational graph - -As dynamic computational model is under develop, We will update more convertable models of dynamic computational graphs as the develop of PaddlePaddle. ## Image classification From 6cbf43398f08455cd6fc4aebf6e801eefae8cf6b Mon Sep 17 00:00:00 2001 From: Jason Date: Tue, 25 May 2021 17:07:45 +0800 Subject: [PATCH 4/5] Update model_zoo.md --- docs/zh/model_zoo.md | 123 +++++++++++++++---------------------------- 1 file changed, 41 insertions(+), 82 deletions(-) diff --git a/docs/zh/model_zoo.md b/docs/zh/model_zoo.md index af351e1d3..78ff1ca5c 100644 --- a/docs/zh/model_zoo.md +++ b/docs/zh/model_zoo.md @@ -1,95 +1,54 @@ -目前paddle2onnx工具集主要支持转化的模型有三大类，图像分类，图像检测和图像分割。随着Paddle 2.0的开发，序列化算子的实现将更加通用，未来有望支持NLP, OCR系列模型的转换。 -受限于不同框架的差异，部分模型可能会存在目前无法转换的情况，如若您发现无法转换或转换失败，或者转换后模型预测存在误差等问题，欢迎通过[ISSUE反馈](https://github.com/PaddlePaddle/paddle-onnx/issues/new)的方式告知我们(模型名，代码实现或模型获取方式)，我们会即时跟进： +Paddle2ONNX支持将飞桨模型转换为ONNX格式存储。 + +由于不同框架的差异，如您在转换过程中出现错误，可随时通过issue向我们反馈，我们的工程师会及时在线回复[ISSUE](https://github.com/PaddlePaddle/paddle-onnx/issues/new)。 -# 动态图模型 -目前paddle动态图模型还在开发中，随着Paddle动态图模型的增加，我们会及时更新可转换的动态图模型类型。 ## 图像分类 -图像分类模型比较完善，目前已支持paddlecls [dygraph分支](https://github.com/paddlepaddle/paddleclas/tree/dygraph)中全系列的模型。 +目前已支持PaddlClas中大部分模型 [release/2.1](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1). -|模型名称 | 来源 | +|Models | Source | |---|---| -| ResNet及其Vd系列 | [paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#resnet%e5%8f%8a%e5%85%b6vd%e7%b3%bb%e5%88%97)| -| 移动端系列(MobileNet等)| [paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#%e7%a7%bb%e5%8a%a8%e7%ab%af%e7%b3%bb%e5%88%97)| -| SEResNeXt与Res2Net系列 | [paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#seresnext%e4%b8%8eres2net%e7%b3%bb%e5%88%97)| -| DPN与DenseNet系列 |[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#dpn%e4%b8%8edensenet%e7%b3%bb%e5%88%97)| -| HRNet系列|[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#hrnet%e7%b3%bb%e5%88%97)| -| Inception系列 |[paddleclas](https://github.com/PaddlePaddle/PaddleClas/blob/dygraph/README_cn.md#inception%E7%B3%BB%E5%88%97)| -| EfficientNet与ResNeXt101_wsl系列 |[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#efficientnet%e4%b8%8eresnext101_wsl%e7%b3%bb%e5%88%97)| -| ResNeSt与RegNet系列|[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#resnest%e4%b8%8eregnet%e7%b3%bb%e5%88%97)| -| Transformer系列 |[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#transformer%e7%b3%bb%e5%88%97)| -| 其他模型 |[paddleclas](https://github.com/paddlepaddle/paddleclas/blob/dygraph/readme_cn.md#%e5%85%b6%e4%bb%96%e6%a8%a1%e5%9e%8b)| - -## 图像OCR -支持的模型有DB(文字检测)，CRNN(文字识别)，以及方向分类模型，测试模型来自于PaddleOCR [dgraph分支](https://github.com/PaddlePaddle/PaddleOCR//tree/dygraph)。 - -| 模型 | 来源 | -|-------|--------| -|DB|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/algorithm_overview.md#1%E6%96%87%E6%9C%AC%E6%A3%80%E6%B5%8B%E7%AE%97%E6%B3%95) | -|CRNN|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/algorithm_overview.md#2%E6%96%87%E6%9C%AC%E8%AF%86%E5%88%AB%E7%AE%97%E6%B3%95) | -|CLS|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/models_list.md#%E4%B8%89%E6%96%87%E6%9C%AC%E6%96%B9%E5%90%91%E5%88%86%E7%B1%BB%E6%A8%A1%E5%9E%8B) | - -## 图像分割 -支持的模型有UNet, HRNet, DeepLab等模型，测试模型来自于PaddleSeg [release/v2.0.0-rc分支](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.0.0-rc)。 - -| 模型 | 来源 | -|-------|--------| -|BiSeNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.0.0-rc/configs/bisenet) | -|DANet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/danet) | -|DeepLabv3|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/deeplabv3) | -|Deeplabv3P |[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/deeplabv3p) | -|FCN|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/fcn) | -|GCNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/gcnet) | -|OCRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/ocrnet) | -|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.0.0-rc/configs/unet) | +| ResNet series| [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#ResNet_and_Vd_series)| +| Mobile series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#Mobile_series)| +| SEResNeXt and Res2Net series | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#SEResNeXt_and_Res2Net_series)| +| DPN and DenseNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#DPN_and_DenseNet_series)| +| HRNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#HRNet_series)| +| Inception series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#Inception_series)| +| EfficientNet and ResNeXt101_wsl series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#EfficientNet_and_ResNeXt101_wsl_series)| +| ResNeSt and RegNet series |[PaddleClas](https://github.com/PaddlePaddle/PaddleClas/tree/release/2.1#ResNeSt_and_RegNet_series)| -## 图像检测 -待测试 - -## 自然语言处理 -目前支持的模型有ERNIE系列模型，测试模型来自于PaddleNLP [2.0-beta 分支](https://github.com/PaddlePaddle/models/tree/release/2.0-beta/PaddleNLP)。 - -| 模型 | 来源 | -|-------|--------| -|ERNIE-1.0|[PaddleNLP](https://github.com/PaddlePaddle/models/blob/develop/PaddleNLP/docs/models.md#paddlenlpmodels) | -|ERNIE-2.0|[PaddleNLP](https://github.com/PaddlePaddle/models/blob/develop/PaddleNLP/docs/models.md#paddlenlpmodels) | - -# 静态图模型 -## 图像分类 -图像分类模型支持比较完善，测试模型来自于 PaddleCls [master 分支](https://github.com/PaddlePaddle/PaddleClas/tree/master)。 - -| 模型 | 来源 | -|-------|--------| -| ResNet | [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/resnet.py) | -| DenseNet | [PaddleClas](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/densenet.py) | -| ShuffleNet | [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/shufflenet_v2.py) | -| MobileNet| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/mobilenet_v3.py) | -| VGG16| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/vgg.py) | -| SE_ResNext50| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/se_resnext.py) | -| InceptionV4| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/inception_v4.py) | -| SE_ResNet50_vd| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/se_resnext_vd.py) | -| SqueezeNet1_0| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/squeezenet.py) | -| Xception41| [PaddleCls](https://github.com/PaddlePaddle/PaddleClas/blob/master/ppcls/modeling/architectures/xception.py) | -## 图像检测 -支持的模型有SSD、YoloV3、FasterRCNN模型，测试模型来自于PaddleDetection [release/0.4分支](https://github.com/PaddlePaddle/Paddledetection/tree/release/0.4)。由于ONNX对检测模型算子支持比较有限，paddle2onnx对检测模型也不能完全支持。后续我们计划增加对其它检测模型的支持，基于ONNX目前对检测模型支持的现状，将会主要集中于一阶段检测模型。 +## OCR +支持PaddleOCR的轻量级和服务端文字识别模型 PaddleOCR [release/2.1](https://github.com/PaddlePaddle/PaddleOCR/tree/release/2.1)。 -| 模型 | 来源 | +| Models | Source | |-------|--------| -|SSD_MobileNet|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#ssd) | -|YoloV3_DarkNet53|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#yolo-v3-%E5%9F%BA%E4%BA%8Epasacl-voc%E6%95%B0%E6%8D%AE%E9%9B%86) | -|YoloV3_ResNet34|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#yolo-v3-%E5%9F%BA%E4%BA%8Epasacl-voc%E6%95%B0%E6%8D%AE%E9%9B%86) | -|YoloV3_MobileNet|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#yolo-v3-%E5%9F%BA%E4%BA%8Epasacl-voc%E6%95%B0%E6%8D%AE%E9%9B%86) | -|FasterRCNN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#faster--mask-r-cnn) | -|FasterRCNN_FPN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#faster--mask-r-cnn) | -|FasterRCNN_FPN_DCN|[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/blob/release/0.4/docs/MODEL_ZOO_cn.md#deformable-%E5%8D%B7%E7%A7%AF%E7%BD%91%E7%BB%9Cv2) | +|Chinese and English ultra-lightweight OCR model (9.4M) |[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR#pp-ocr-20-series-model-listupdate-on-dec-15) | +|Chinese and English general OCR model (143.4M)|[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR#pp-ocr-20-series-model-listupdate-on-dec-15) | -## 图像分割 -支持的模型有UNet, HRNet, DeepLab模型，测试模型来自于PaddleSeg [release/v0.7.0分支](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v0.7.0) +## 语义分割 +支持语义分割模型库PaddleSeg中的大部分模型 [release/v2.1](https://github.com/PaddlePaddle/PaddleSeg/tree/release/2.1)。 -| 模型 | 来源 | +| Models | Source | |-------|--------| -|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_unet.md) | -|HRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_hrnet.md) | -|DeepLab|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v0.7.0/tutorial/finetune_deeplabv3plus.md) | +|BiSeNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/tree/release/v2.1/configs/bisenet) | +|DANet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/danet) | +|DeepLabv3|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/deeplabv3) | +|Deeplabv3P |[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/deeplabv3p) | +|FCN|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/fcn) | +|GCNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/gcnet) | +|OCRNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/ocrnet) | +|UNet|[PaddleSeg](https://github.com/PaddlePaddle/PaddleSeg/blob/release/v2.1/configs/unet) | + +## 目标检测 +支持目标检测模型库中8种检测结构 [release/2.1](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.1) +| Models | Source | +| ----------- | ------------------------------------------------------------ | +| YOLO-V3 | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/yolov3/ | +| PPYOLO | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| PPYOLO-Tiny | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| PPYOLO-V2 | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyolo/ | +| TTFNet | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ttfnet/ | +| PAFNet | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ttfnet/ | +| SSD | https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ssd/ | From 7928b3e269e39b2f3ab6effda51d3751d8599ec1 Mon Sep 17 00:00:00 2001 From: jiangjiajun Date: Tue, 1 Jun 2021 09:57:34 +0000 Subject: [PATCH 5/5] Add pixel_shuffle support --- paddle2onnx/op_mapper/tensor.py | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/paddle2onnx/op_mapper/tensor.py b/paddle2onnx/op_mapper/tensor.py index 6db5cc8f5..dbdd0093f 100644 --- a/paddle2onnx/op_mapper/tensor.py +++ b/paddle2onnx/op_mapper/tensor.py @@ -1078,3 +1078,19 @@ def compute_output_shape_by_size(cls, graph, node, opset_version=10): shape_node3 = graph.make_node( 'Concat', inputs=[shape_node1, shape_node2], axis=0) return shape_node0, shape_node3 + +@op_mapper('pixel_shuffle') +class PixelShuffle(): + support_opset_verison_range = (11, 12) + + @classmethod + def opset_11(cls, graph, node, **kw): + upscale_factor = node.attr('upscale_factor') + + node = graph.make_node( + 'DepthToSpace', + inputs=node.input('X'), + outputs=node.output('Out'), + blocksize=upscale_factor, + mode='CRD') +