ERROR: cv::dnn::dnn4_v20211004::LayerData&) () from /usr/local/lib/libopencv_world.so.4.5

[karishmathumu]

0x0000ffffee8cb194 in cv::dnn::dnn4_v20211004::Net::Impl::forwardLayer(cv::dnn::dnn4_v20211004::LayerData&) () from /usr/local/lib/libopencv_world.so.4.5

**

• OpenCV => 4.5.4
• Operating System / Platform => Linux : aarch64-gnu , Ubuntu 20.04
• Compiler => Visual Studio 2019
• DARKNET
• CUDA
• JETSON ORIN
• **

I am trying to implement R-CNN Masking on my custom object detection model with yolov3 - DAKRNET and opencv 4.5.4 on Jetson Orin.

I have succeeded in the Object detection with CUDA backend support

Now in the masking implementation for the same, I followed this tutorial in cpp
https://github.com/spmallick/learnopencv/blob/master/Mask-RCNN/mask_rcnn.cpp

String textGraph = "./mask_rcnn_inception_v2_coco_2018_01_28.pbtxt";
String modelWeights = "./mask_rcnn_inception_v2_coco_2018_01_28/frozen_inference_graph.pb";

    // Load the network
    Net net = readNetFromTensorflow(modelWeights, textGraph);

    if (device == "cpu")
    {
        cout << "Using CPU device" << endl;
        net.setPreferableBackend(DNN_TARGET_CPU);
    }
    else if (device == "gpu")
    {
        cout << "Using GPU device" << endl;
        net.setPreferableBackend(DNN_BACKEND_CUDA);
        net.setPreferableTarget(DNN_TARGET_CUDA);
    }

I replaced the respective above lines of this code with the following lines

String modelConfig = "/home/.../darknet/cfg/yolo-obj-3classes.cfg";
String modelWts = "/home/.../darknet/yolo-obj_final_3classes.weights";

    // Load the network
Net net = readNetFromDarknet(modelConfig, modelWts);
net.setPreferableBackend(DNN_BACKEND_OPENCV);
net.setPreferableTarget(DNN_TARGET_CPU);

P.S: while the exact same lines worked prefectly good with the plain object detection

And my error when I run this on the Visual Studio 2019 is as follows:

0x0000ffffee8cb194 in cv::dnn::dnn4_v20211004::Net::Impl::forwardLayer(cv::dnn::dnn4_v20211004::LayerData&) () from /usr/local/lib/libopencv_world.so.4.5

Seeking for help and suggestions.

Kind regards,
Karishma

[krrishdholakia]

Are you using the same versions of CUDA and darknet as in the tutorial?

[karishmathumu]

Yes, I did not find in the mentioned tutorial which CUDA version and darknet it uses.

But my combination of CUDA 11.4 and AlexeyAB - DARKNET for this tutorial's predecesser tutorial which is https://learnopencv.com/deep-learning-based-object-detection-using-yolov3-with-opencv-python-c/

and I used my custom-trained yolov3 model - a) weights and b) config file for the object detection.

So, continuing with those very same weights and config files in this tutorial for Instance segmentation.

[asmorkalov]

@krrishdholakia Is it still relevant for OpenCV 4.8.0?