No Caffe Darknet frameworks in opencv 5.0

1

Darknet importer remove in opencv 5.0

Caffe importer remove in opencv 5.0

2

If you need caffe or darknet in opencv 5 you can copy this repo and move it in opencv_contrib before running cmake

3

and if you need to use an onnx model with dynamic input you can use dnnlegacy::importOnnxWithFixedShape function

paper GitHub - xavysp/DexiNed: DexiNed: Dense EXtreme Inception Network for Edge Detection
model https://storage.googleapis.com/ailia-models/dexined/model.onnx

image
image1288×728 33.6 KB

function dnnlegacy::importOnnxWithFixedShape repalce w and h by 640 and 480

        std::map<std::string, int> val;
        val["w"] = 640;
        val["h"] = 480;
        cv::dnn::Net x = dnnlegacy::importOnnxWithFixedShape("dexined.onnx", val);
        Mat img = imread(samples::findFile("butterfly.jpg"));
        Image2BlobParams paramBlob;
        paramBlob.datalayout = DNN_LAYOUT_NCHW;
        paramBlob.ddepth = CV_32F;
        paramBlob.mean = Scalar(0, 0, 0, 0);
        paramBlob.scalefactor = Scalar(1, 1, 1, 1);
        paramBlob.size = Size(640, 480);
        paramBlob.swapRB = false;
        paramBlob.paddingmode = DNN_PMODE_NULL;
        Mat blobNHWC = blobFromImageWithParams(img, paramBlob);
        x.setInput(blobNHWC);
        Mat y = x.forward("block_cat");