How to convert Yolov3-Resnet18 from ModelArts into Atlas 200 DK and use it for video object detection?

windyxyz · Created Apr 25, 2021 01:12:10

Hi， the error is ”fatal error: acl/acl.h: No such file or directory“
You may check src/Cmakelists.txt, make sure whether “include_directories” include acl/acl.h directory ? If not add it 。

include_directories(
$ENV{HOME}/ascend_ddk/${ARCH}/include/
$ENV{HOME}/ascend_ddk/${ARCH}/include/opencv4
${INC_PATH}/acllib/include/
../inc/

ariase88 · Created Apr 23, 2021 17:23:18

Thanks for contacting the Huawei community!

We are checking your question and will provide an answer to you shortly...

windyxyz · Created Apr 25, 2021 01:12:10

Hi， the error is ”fatal error: acl/acl.h: No such file or directory“
You may check src/Cmakelists.txt, make sure whether “include_directories” include acl/acl.h directory ? If not add it 。

include_directories(
$ENV{HOME}/ascend_ddk/${ARCH}/include/
$ENV{HOME}/ascend_ddk/${ARCH}/include/opencv4
${INC_PATH}/acllib/include/
../inc/

sliawatimena · Created Apr 26, 2021 07:11:29

At src/CMakeList.txt, I add line 47

Why I cannot select Target OS and Target Architecture at Build Configuration?

build

However, I still have src/.build_project

Is the .build_project correct?

I try to reinstall MindStudio 2.0.0 (beta)

$ tar -zxvf $HOME/Downloads/MS200/MindStudio_2.0.0-beta3_linux.tar.gz

But, it still give blank Target OS and Target Architecture.

I open another project such as

/home/suryadi/samples/cplusplus/level2_simple_inference/1_classification/googlenet_imagenet_picture/

But, it still give blank Target OS and Target Architecture.

I think this export DDK_PATH should be set to

DDK_PATH= /home/suryadi/Ascend/ascend-toolkit/latest/arm64-linux

How to set this build command?

Information:Build command:

Information:

export DDK_PATH=/home/suryadi/Ascend/ascend-toolkit/20.2.rc1/acllib &&

export NPU_HOST_LIB=/home/suryadi/Ascend/ascend-toolkit/20.2.rc1/acllib/lib64/stub &&

export lang="C" &&

cd /home/suryadi/samples/cplusplus/level2_simple_inference/2_object_detection/YOLOV3_coco_detection_video/build/intermediates &&

rm -rf * &&

cmake /home/suryadi/samples/cplusplus/level2_simple_inference/2_object_detection/YOLOV3_coco_detection_video -DCMAKE_BUILD_TYPE=Debug -DCMAKE_SKIP_RPATH=TRUE -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++ &&

make &&

echo YOLOV3_coco_detection_video build successfully.

The errors

eer

Information:[ 80%] Building CXX object src/CMakeFiles/main.dir/main.cpp.o

Information:[100%] Linking CXX executable ../../../out/main

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: skipping incompatible /home/suryadi/Ascend/ascend-toolkit/20.2.rc1/acllib/lib64/stub/libascendcl.so when searching for -lascendcl

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: cannot find -lascendcl

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: skipping incompatible /home/suryadi/Ascend/ascend-toolkit/20.2.rc1/acllib/lib64/stub/libacl_dvpp.so when searching for -lacl_dvpp

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: cannot find -lacl_dvpp

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: skipping incompatible /usr/local/lib/libprotobuf.so when searching for -lprotobuf

Information:/usr/lib/gcc-cross/aarch64-linux-gnu/7/../../../../aarch64-linux-gnu/bin/ld: skipping incompatible /usr/local/lib/libprotobuf.a when searching for -lprotobuf

Error:collect2: error: ld returned 1 exit status

Error:src/CMakeFiles/main.dir/build.make:147: recipe for target '../../out/main' failed

Error:make[2]: *** [../../out/main] Error 1

Error:CMakeFiles/Makefile2:112: recipe for target 'src/CMakeFiles/main.dir/all' failed

Error:make[1]: *** [src/CMakeFiles/main.dir/all] Error 2

Error:Makefile:148: recipe for target 'all' failed

Error:make: *** [all] Error 2

How to solve this problem?

Thank you very much in advance.

Warmest Regards,

Suryadi

windyxyz · Created Apr 28, 2021 02:06:03

Hi，It seems that you didn‘t configure MindStudio toolkit ？

You can refer to：

https://support.huaweicloud.com/intl/en-us/usermanual-mindstudio301/atlasms_02_0039.html

sliawatimena · Created May 11, 2021 13:07:22

Dear all,

I can run Model Converter inside MindStudio 3.0.1 and try to convert yolo3_resnet

The ATC run success.

However, I try to run the YOLOV3_VOC_detection_picture program

[INFO] load model ../model/yolo3_resnet18.om success

modelDesc_ = 0xaaaab688e740

modelId_ = /strong>

I change kModelHeight = 352 and kModelWidth = 640

The following is the yolo3_resnet18_config.json content:

{"convertConfig":{"advancedArgs":"--output_type=FP32 --save_original_model=false --log=info","aippBtnSelection":[{"aippButtonName":"images","aippCropBtnSelected":false,"aippMeanBtnSelected":true,"aippMinBtnSelected":false,"aippMultiplierBtnSelected":false,"aippOpenBtSelected":true,"aippPaddingBtnSelected":false,"aippPreImgBtnSelected":true,"aippResizeBtnSelected":false}],"aippData":[{"aippMode":"STATIC","bottomPaddingSize":"","crop":false,"cropH":"","cropStartX":"","cropStartY":"","cropW":"","enable":true,"inputImageFormat":"BGR888_U8","inputImageFormatShowValue":"BGR package","inputImageSizeH":"352","inputImageSizeW":"640","inputIndex":"0","leftPaddingSize":"","maxImageSize":"","meanLess0":"104","meanLess1":"117","meanLess2":"123","meanLess3":"","minChn0":"0.0","minChn1":"0.0","minChn2":"0.0","minChn3":"","modelImageFormat":"RGB","modelImageType":"BT_601NARROW","multiFactor0":"1.0","multiFactor1":"1.0","multiFactor2":"1.0","multiFactor3":"","padding":false,"resize":false,"resizeOutputH":"","resizeOutputW":"","rightPaddingSize":"","topPaddingSize":""}],"aippMode":"static","aippSelection":true,"autoTuneBtnSeleted":false,"autoTuneMode":"","batchList":[],"batchNum":"0","calData":[],"closeFusionRule":"false","customOpsPath":"","dynamicBatch":false,"dynamicBatchContent":"","dynamicImgSize":false,"dynamicImgSizeContent":"","encryptBtnSelection":true,"history":false,"historyItemName":"","inputFormat":"NHWC","inputFormatChooseEnable":true,"inputShape":[{"name":"images","number":"1","channel":"3","width":"640","height":"352","inputShape":"1,352,640,3","inputIndex":"0","inputType":"FP32","addByUser":false,"imagesFile":""}],"inputType":"FP32","inputTypeChanged":false,"markedNodes":[],"modelFilePath":"/home/suryadi/AscendProjects/output_Yolov3-Resnet18/frozen_graph/yolo3_resnet18.pb","modelName":"yolo3_resnet18","modelType":"TENSORFLOW","needAdvancedArgs":true,"outNodes":[],"outNodesItem":[],"outputFolderPath":"/home/suryadi/modelzoo/yolo3_resnet18","outputSoc":"Ascend310","outputType":"FP32","outputTypeSwitch":false,"pluginPath":"","quantParamFilePath":"","targetType":"DEVICE","weightFilePath":""},"name":"yolo3_resnet18_config"}

However, it stuck at execture model failed, modelId is 1.

How to resolve this problem?

Thank you very much in advance.

Warmest Regards,

Suryadi

windyxyz · Created May 12, 2021 00:55:16

Hi， your model is different with the sample which you use。 please refer to this sample：
https://gitee.com/ascend/samples/tree/master/python/level2_simple_inference/2_object_detection/YOLOV3_mask_detection_picture

How to convert Yolov3-Resnet18 from ModelArts into Atlas 200 DK and use it for video object detection?

Recommended answer

Third Party’s Trade Secret

My Followers

Enterprise

Consumer

Corporate

Carriers

How to convert Yolov3-Resnet18 from ModelArts into Atlas 200 DK and use it for video object detection?

Recommended answer

Third Party’s Trade Secret

My Followers