Platforms:
- Win10 x64, CMake 3.13.4, VS2017, OpenCV build with MKL, TBB, eigen, OpenVINO 2019.2.242, opencv rev. c95407f.
- Xubuntu 18.04.1, CMake, OpenCV build with MKL, TBB, eigen, OpenVINO 2019.1.144, OpenCV rev. f6ec0cd in Virtual Box.
CPU`s dont support AVX++ instruction.
Сode:
Mat Img = imread("e:/q/Avatar_cat.png");
Net BaseNet = readNet("O:/0/object_detection/common/faster_rcnn/faster_rcnn_resnet101_coco/tf/faster_rcnn_resnet101_coco_2018_01_28/frozen_inference_graph.pb",
"O:/0/object_detection/common/faster_rcnn/faster_rcnn_resnet101_coco/tf/faster_rcnn_resnet101_coco_2018_01_28/faster_rcnn_resnet101_coco_2018_01_28.pbtxt");
BaseNet.setPreferableBackend(DNN_BACKEND_INFERENCE_ENGINE);//with DNN_BACKEND_OPENCV work fine
BaseNet.setPreferableTarget(DNN_TARGET_CPU);
Mat blob;
std::vector<Mat> outs;
try {
blobFromImage(Img, blob);
BaseNet.setInput(blob);
std::vector<String> outNames = BaseNet.getUnconnectedOutLayersNames();
BaseNet.forward(outs, outNames); //exception here
}
catch (const std::exception &e) {
cout << e.what();
return -1;
}
On Net forward with Inference engine backend received error:
OpenCV(4.1.1-dev) Error: Assertion failed (> Failed to initialize Inference Engine backend: data [detection_out/slice] doesn't exist
> ..\src\inference_engine\cnn_network_impl.cpp:173
> e:\lib_prebuild\openvino\openvino\inference_engine\include\details/ie_exception_conversion.hpp:71
) in cv::dnn::InfEngineBackendNet::initPlugin, file E:\Lib_prebuild\opencv\source\opencv\modules\dnn\src\op_inf_engine.cpp, line 477
OpenCV(4.1.1-dev) E:\Lib_prebuild\opencv\source\opencv\modules\dnn\src\op_inf_engine.cpp:477: error: (-215:Assertion failed) in function 'cv::dnn::InfEngineBackendNet::initPlugin'
> Failed to initialize Inference Engine backend: data [detection_out/slice] doesn't exist
> ..\src\inference_engine\cnn_network_impl.cpp:173
> e:\lib_prebuild\openvino\openvino\inference_engine\include\details/ie_exception_conversion.hpp:71
With OpenCV backend all works fine.
Models faster_rcnn_resnet101_coco_2018_01_28 downloaded by OpenVINO model downloader. Proto file from here https://gist.github.com/dkurt/1fe03728dcb51955c89d30dfe6c11a3d#file-faster_rcnn_resnet101_coco_2018_01_28-pbtxt.
Platforms:
CPU`s dont support AVX++ instruction.
Сode:
On
Net forwardwith Inference engine backend received error:With OpenCV backend all works fine.
Models faster_rcnn_resnet101_coco_2018_01_28 downloaded by OpenVINO model downloader. Proto file from here https://gist.github.com/dkurt/1fe03728dcb51955c89d30dfe6c11a3d#file-faster_rcnn_resnet101_coco_2018_01_28-pbtxt.