From c97884b6ad388c39b9e9da460d6e1f2dda07e9b0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=BD=AD=E4=B8=9A=E5=BA=86?= Date: Mon, 26 Jul 2021 07:53:17 +0000 Subject: [PATCH 1/5] =?UTF-8?q?update=20Ascend-PyTorch=E7=A6=BB=E7=BA=BF?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/ONNX=E6=A8=A1=E5=9E=8B?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/official/cv/ReID/ReID-stro?= =?UTF-8?q?ng-baseline/README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../official/cv/ReID/ReID-strong-baseline/README.md" | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" index 2329caf..7cbb063 100644 --- "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" +++ "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" @@ -43,8 +43,8 @@ bash test/eval_acc_perf.sh --datasets_path=/root/datasets **评测结果：** | 模型 | 官网pth精度 | 310离线推理精度 | 基准性能 | 310性能 | | :------: | :------: | :------: | :------: | :------: | -| ReID-strong-baseline bs1 | [rank1:94.5% mAP:85.9%](https://github.com/michuanhaohao/reid-strong-baseline) | rank1:94.5% mAP:85.9% | 992.9994fps | 1446.188fps | -| ReID-strong-baseline bs16 | [rank1:94.5% mAP:85.9%](https://github.com/michuanhaohao/reid-strong-baseline) | rank1:94.5% mAP:85.9% | 2211.7074fps | 2588.56fps | +| ReID-strong-baseline bs1 | [rank1:94.5% mAP:85.9%](https://github.com/michuanhaohao/reid-strong-baseline) | rank1:94.5% mAP:85.9% | 995fps | 1446.188fps | +| ReID-strong-baseline bs16 | [rank1:94.5% mAP:85.9%](https://github.com/michuanhaohao/reid-strong-baseline) | rank1:94.5% mAP:85.9% | 2210fps | 2588.56fps | 备注： 加上TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx可以进行离线推理 -- Gitee From 373860838fe1408926b3f1e0ad543e06c833aaca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=BD=AD=E4=B8=9A=E5=BA=86?= Date: Mon, 26 Jul 2021 07:57:04 +0000 Subject: [PATCH 2/5] =?UTF-8?q?update=20Ascend-PyTorch=E7=A6=BB=E7=BA=BF?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/ONNX=E6=A8=A1=E5=9E=8B?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/official/cv/ReID/ReID-stro?= =?UTF-8?q?ng-baseline/README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../official/cv/ReID/ReID-strong-baseline/README.md" | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" index 7cbb063..3de315e 100644 --- "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" +++ "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" @@ -8,7 +8,7 @@ $\color{red}{说明：删除线用于READ.md的说明，以下带有删除线的 ``` pip3.7 install -r requirements.txt ``` -~~需要使用python3.7命令执行脚本，pip3.7命令安装库，torch使用1.5.0版本，如果开源模型代码导出onnx要求torch版本大于1.5.0，则使用1.8.0版本，并在此处说明。onnx默认选择1.7.0。requirements.txt中需要写明本模型离线推理所有必要依赖库的具体版本，版本号即是推理310服务器上推理时使用库的版本号，常用库包括numpy，Pillow，opencv-python等。目前atc工具支持的onnx算子版本opset_version为11。~~ +~~需要使用python3.7命令执行脚本，pip3.7命令安装库，torch使用开源1.5.0版本，如果开源模型代码导出onnx要求torch版本大于1.5.0，则使用1.8.0版本，并在此处说明。onnx默认选择1.7.0。requirements.txt中需要写明本模型离线推理所有必要依赖库的具体版本，版本号即是推理310服务器上推理时使用库的版本号，常用库包括numpy，Pillow，opencv-python等。目前atc工具支持的onnx算子版本opset_version为11。~~ 2.获取，修改与安装开源模型代码 @@ -25,7 +25,7 @@ cd .. 3.获取权重文件 [market_resnet50_model_120_rank1_945.pth](https://drive.google.com/open?id=1hn0sXLZ5yJcxtmuY-ItQfYD7hBtHwt7A) -~~优先使用训练提供的权重文件，如果训练的权重文件网上能获则需给出网址，否则需要给出从哪获取权重文件。如果训练没有提供权重则使用开源代码仓的权重文件。需要给出权重文件名~~ +~~优先使用训练提供的权重文件，如果训练的权重文件网上能获则需给出网址，否则需要给出从哪获取权重文件。如果训练没有提供权重则使用开源代码仓基础配置的权重文件。需要给出权重文件名~~ 4.数据集 [获取Market1501](http://www.liangzheng.org/Project/project_reid.html)，并重命名为market1501 -- Gitee From 0f45cbaa0e16da272cbd2bd4739750f366ef1a4e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=BD=AD=E4=B8=9A=E5=BA=86?= Date: Mon, 26 Jul 2021 08:21:37 +0000 Subject: [PATCH 3/5] =?UTF-8?q?update=20Ascend-PyTorch=E7=A6=BB=E7=BA=BF?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/ONNX=E6=A8=A1=E5=9E=8B?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/official/cv/ReID/ReID-stro?= =?UTF-8?q?ng-baseline/README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../official/cv/ReID/ReID-strong-baseline/README.md" | 1 + 1 file changed, 1 insertion(+) diff --git "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" index 3de315e..655cc04 100644 --- "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" +++ "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" @@ -49,5 +49,6 @@ bash test/eval_acc_perf.sh --datasets_path=/root/datasets 备注： 加上TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx可以进行离线推理 不加TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx转换的om精度与官网精度一致 +~~模型不支持batch 16等特殊情况，需要备注一下~~ -- Gitee From d3d774629a5b52146f405427706bdb638e7e54f6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=BD=AD=E4=B8=9A=E5=BA=86?= Date: Mon, 26 Jul 2021 08:22:51 +0000 Subject: [PATCH 4/5] =?UTF-8?q?update=20Ascend-PyTorch=E7=A6=BB=E7=BA=BF?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/ONNX=E6=A8=A1=E5=9E=8B?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/official/cv/ReID/ReID-stro?= =?UTF-8?q?ng-baseline/README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../official/cv/ReID/ReID-strong-baseline/README.md" | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" index 655cc04..8e60d15 100644 --- "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" +++ "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" @@ -49,6 +49,6 @@ bash test/eval_acc_perf.sh --datasets_path=/root/datasets 备注： 加上TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx可以进行离线推理 不加TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx转换的om精度与官网精度一致 -~~模型不支持batch 16等特殊情况，需要备注一下~~ +~~onnx不支持动态batch，模型不支持batch 16等特殊情况，需要备注一下~~ -- Gitee From 85c53ab69042d689f8757d3e8a4667d9c09cd38b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=BD=AD=E4=B8=9A=E5=BA=86?= Date: Mon, 26 Jul 2021 08:24:12 +0000 Subject: [PATCH 5/5] =?UTF-8?q?update=20Ascend-PyTorch=E7=A6=BB=E7=BA=BF?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/ONNX=E6=A8=A1=E5=9E=8B?= =?UTF-8?q?=E6=8E=A8=E7=90=86=E6=8C=87=E5=AF=BC/official/cv/ReID/ReID-stro?= =?UTF-8?q?ng-baseline/README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../official/cv/ReID/ReID-strong-baseline/README.md" | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" index 8e60d15..1e6369f 100644 --- "a/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" +++ "b/Ascend-PyTorch\347\246\273\347\272\277\346\216\250\347\220\206\346\214\207\345\257\274/ONNX\346\250\241\345\236\213\346\216\250\347\220\206\346\214\207\345\257\274/official/cv/ReID/ReID-strong-baseline/README.md" @@ -49,6 +49,6 @@ bash test/eval_acc_perf.sh --datasets_path=/root/datasets 备注： 加上TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx可以进行离线推理 不加TEST.NECK_FEAT "('before')" TEST.FEAT_NORM "('no')"导出的onnx转换的om精度与官网精度一致 -~~onnx不支持动态batch，模型不支持batch 16等特殊情况，需要备注一下~~ +~~onnx不支持动态batch，onnx因包含npu自定义算子而不能推理，模型不支持batch 16等特殊情况，需要备注一下~~ -- Gitee