From 21852aba88f14163be3559f9d8c89f5c799ae4eb Mon Sep 17 00:00:00 2001 From: lifangtian Date: Thu, 5 Jun 2025 02:01:28 +0000 Subject: [PATCH 1/4] =?UTF-8?q?update=20ACL=5FPyTorch/built-in/cv/FCENet?= =?UTF-8?q?=5Ffor=5FPyTorch/readme.md=EF=BC=9A=E5=A2=9E=E5=8A=A0=E8=8E=B7?= =?UTF-8?q?=E5=8F=96modelzoo=E6=BA=90=E7=A0=81=E4=BB=93=E7=8E=AF=E8=8A=82?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: lifangtian --- ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md b/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md index 22b91e6c32..212aaa6498 100644 --- a/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md +++ b/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md @@ -58,6 +58,13 @@ FCENet，使用傅里叶变换来得到文本的包围框，该方法在弯曲 # 快速上手 ## 安装 +- 获取本仓源码 + + ``` + git clone https://gitee.com/ascend/ModelZoo-PyTorch.git + cd ModelZoo-PyTorch/ACL_PyTorch/contrib/cv/detection/FCENet + ``` + - 安装推理过程所需的依赖 ```bash -- Gitee From e48e8c7fca59fcc79309749ea5e7bc876112c9f0 Mon Sep 17 00:00:00 2001 From: lifangtian Date: Thu, 5 Jun 2025 02:02:18 +0000 Subject: [PATCH 2/4] =?UTF-8?q?update=20ACL=5FPyTorch/built-in/cv/FCENet?= =?UTF-8?q?=5Ffor=5FPyTorch/readme.md=EF=BC=9A=E5=A2=9E=E5=8A=A0=E8=8E=B7?= =?UTF-8?q?=E5=8F=96modelzoo=E6=BA=90=E7=A0=81=E4=BB=93=E7=8E=AF=E8=8A=82?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: lifangtian --- ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md b/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md index 212aaa6498..b383490b9e 100644 --- a/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md +++ b/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch/readme.md @@ -62,7 +62,7 @@ FCENet，使用傅里叶变换来得到文本的包围框，该方法在弯曲 ``` git clone https://gitee.com/ascend/ModelZoo-PyTorch.git - cd ModelZoo-PyTorch/ACL_PyTorch/contrib/cv/detection/FCENet + cd ModelZoo-PyTorch/ACL_PyTorch/built-in/cv/FCENet_for_PyTorch ``` -- Gitee From 4f1dacd6fbf93b18fa3aef72a15efd88ff2df2b0 Mon Sep 17 00:00:00 2001 From: lifangtian Date: Thu, 5 Jun 2025 02:48:27 +0000 Subject: [PATCH 3/4] =?UTF-8?q?update=20ACL=5FPyTorch/built-in/cv/PSENet?= =?UTF-8?q?=5FResNet=5Ffor=5FPP/README.md=EF=BC=9A=E8=A1=A5=E5=85=85?= =?UTF-8?q?=E6=95=B0=E6=8D=AE=E9=9B=86=E7=9A=84=E9=93=BE=E6=8E=A5=EF=BC=9B?= =?UTF-8?q?=E6=9B=B4=E6=AD=A3=E8=A7=A3=E5=8E=8B=E5=91=BD=E4=BB=A4=EF=BC=9B?= =?UTF-8?q?=E6=9B=BF=E6=8D=A2=E5=A4=B1=E6=95=88=E7=9A=84auto-optimizer?= =?UTF-8?q?=E9=93=BE=E6=8E=A5?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: lifangtian --- .../built-in/cv/PSENet_ResNet_for_PP/README.md | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) diff --git a/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/README.md b/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/README.md index 31fcaba899..dde2be267b 100644 --- a/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/README.md +++ b/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/README.md @@ -64,6 +64,13 @@ PSENet([Shape Robust Text Detection with Progressive Scale Expansion Network](ht # 快速上手 ## 获取源码 +1. 获取本仓源码 + + ``` + git clone https://gitee.com/ascend/ModelZoo-PyTorch.git + cd ModelZoo-PyTorch/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP + ``` + 1. 获取源码。 @@ -89,12 +96,12 @@ PSENet([Shape Robust Text Detection with Progressive Scale Expansion Network](ht 1. 获取原始数据集。（解压命令参考tar –xvf \*.tar与 unzip \*.zip） - IICDAR 2015 数据集包含1000张训练图像和500张测试图像。参考PaddleOCR数据集数据处理方式，ICDAR 2015 数据集可以点击链接进行下载，本模型需下载Test Set Images(43.3MB)。 + ICDAR 2015 数据集包含1000张训练图像和500张测试图像。参考PaddleOCR数据集数据处理方式，ICDAR 2015 数据集可以点击[链接](https://aistudio.baidu.com/datasetdetail/46088)进行下载，本模型需下载Test Set Images(43.3MB)。 将数据集`ch4_test_images.zip`放在工作目录下，通过以下命令创建`train_data/icdar2015/ch4_test_images`路径，并通过以下命令进行解压保存并获取标签文件。 ``` mkdir -p ./train_data/icdar2015/ch4_test_images/ - unzip -d ./train_data/icdar2015/ch4_test_images/ ch4_test_images.zip + unzip -d ch4_test_images.zip ./train_data/icdar2015/ch4_test_images/ wget -P ./train_data/icdar2015/ https://paddleocr.bj.bcebos.com/dataset/test_icdar2015_label.txt ``` @@ -166,12 +173,12 @@ PSENet([Shape Robust Text Detection with Progressive Scale Expansion Network](ht 运行后获得PSENet_ResNet.onnx文件。 2. 优化onnx模型。 - 请访问[auto-optimizer优化工具](https://gitee.com/ascend/msadvisor/tree/master/auto-optimizer)代码仓，根据readme文档进行工具安装。 + 请访问[auto-optimizer优化工具](https://gitee.com/ascend/msit/blob/master/msit/README.md)代码仓，根据readme文档进行benchmark和surgeon工具安装。 运行modify_onnx.py脚本优化onnx模型，优化点为：Resize算子按Paddle定义参数导出的onnx模型有精度问题，因此将按PyTorch定义重新构造Resize参数。 ``` - python3 modify_onnx.py PSENet_ResNet PSENet_ResNet_md_dybs.onnx + python3 modify_onnx.py PSENet_ResNet.onnx PSENet_ResNet_md_dybs.onnx ``` 参数说明:第一个为输入onnx，第二个为输出onnx。 -- Gitee From b8215226caca25ec95892ee57cd3ada4c2ffc3d5 Mon Sep 17 00:00:00 2001 From: lifangtian Date: Thu, 5 Jun 2025 02:50:14 +0000 Subject: [PATCH 4/4] =?UTF-8?q?update=20ACL=5FPyTorch/built-in/cv/PSENet?= =?UTF-8?q?=5FResNet=5Ffor=5FPP/requirements.txt=EF=BC=9A=E5=A2=9E?= =?UTF-8?q?=E5=8A=A0=E9=9C=80=E8=A6=81=E7=9A=84onnxsim=E4=BE=9D=E8=B5=96?= =?UTF-8?q?=EF=BC=9B=E9=99=8D=E7=BA=A7protobuf=E4=BB=A5=E8=B7=91=E9=80=9A?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: lifangtian --- ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/requirements.txt | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/requirements.txt b/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/requirements.txt index 7f8f91e094..e7bacf214f 100644 --- a/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/requirements.txt +++ b/ACL_PyTorch/built-in/cv/PSENet_ResNet_for_PP/requirements.txt @@ -19,4 +19,6 @@ lxml premailer openpyxl attrdict -sympy \ No newline at end of file +sympy +onnxsim +protobuf==3.20.3 \ No newline at end of file -- Gitee