Java happens problem when using Dnn.readFromDarknet with yolov3

Android/Java
1

Environment: Windows 10 Pro
OpenCV: 4.5.1, I downloaded from https://opencv.org/releases/
yolov3.cfg: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3.cfg

yolov3.weights: https://pjreddie.com/media/files/yolov3.weights
yolov3-tiny.cfg: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-tiny.cfg
yolov3-tiny.weights: https://pjreddie.com/media/files/yolov3-tiny.weights
coco.names: https://github.com/AlexeyAB/darknet/raw/master/data/coco.names

My code:

static final Scalar scalar =  new Scalar(0);
static final Size sz = new Size(13*16*4,13*9*4); // new Size(416,416);
static final Size cfgSize = new Size(608, 608);
static final float scale =  1F/255F;
static final boolean swapRB = false;

private static List<String> getOutputNames(Net net) {
    List<String> names = new ArrayList<>();
    List<Integer> outLayers = net.getUnconnectedOutLayers().toList();
    List<String> layersNames = net.getLayerNames();
    outLayers.forEach((item) -> names.add(layersNames.get(item - 1)));
    return names;
}

public static  void test1() {
    System.loadLibrary(org.opencv.core.Core.NATIVE_LIBRARY_NAME);
    System.loadLibrary("opencv_videoio_ffmpeg451_64");
    List<Mat> result = new ArrayList<>();
    String cfgFile     = "E:/yolo/yolo-coco/yolov3.cfg";
    String weightsFile = "E:/yolo/yolo-coco/yolov3.weights";
    String labelFils   = "E:/yolo/yolo-coco/coco.names";
    String videoFile   = "E:/AV/Vehicles/video_SECTION-1.mp4";
    Net net = Dnn.readNetFromDarknet(cfgFile, weightsFile);
    List<String> outputLayers = getOutputNames(net);
    log.trace("outputLayers:{}", outputLayers);
    VideoCapture vs = new VideoCapture(videoFile);
    log.trace("VideoCapture is opened? {}", vs.isOpened());
    Mat frame = new Mat();
    boolean readOk = vs.read(frame);
    log.trace("readOk:{}, frame size:{}", readOk, frame.size());
    Mat inputBlob = Dnn.blobFromImage(frame, scale, sz, scalar, swapRB, false);
    net.setInput(inputBlob);
    log.debug("CHECK POINT 00001");

    net.forward(result, outputLayers); // outBlobNames:[yolo_82, yolo_94, yolo_106]
    log.trace("Aftert blobFromImage ");
}

Output:

43:58.278 INFO OpencvTest: Loading opencv_java451
14:43:58.529 TRACE OpencvTest: outputLayers:[yolo_82, yolo_94, yolo_106]
14:43:58.552 TRACE OpencvTest: VideoCapture is opened? true
14:43:58.574 TRACE OpencvTest: readOk:true, frame size:1280x720
14:43:58.581 DEBUG OpencvTest: CHECK POINT 00001

Errors:

[ERROR:0] global C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\dnn.cpp (3498) cv::dnn::dnn4_v20201117::Net::Impl::getLayerShapesRecursively OPENCV/DNN: [Concat]:(concat_98): getMemoryShapes() throws exception. inputs=2 outputs=1/1 blobs=0
[ERROR:0] global C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\dnn.cpp (3501) cv::dnn::dnn4_v20201117::Net::Impl::getLayerShapesRecursively input[0] = [ 1 128 60 104 ]
[ERROR:0] global C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\dnn.cpp (3501) cv::dnn::dnn4_v20201117::Net::Impl::getLayerShapesRecursively input[1] = [ 1 256 59 104 ]
[ERROR:0] global C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\dnn.cpp (3505) cv::dnn::dnn4_v20201117::Net::Impl::getLayerShapesRecursively output[0] = [ 1 128 60 104 ]
[ERROR:0] global C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\dnn.cpp (3511) cv::dnn::dnn4_v20201117::Net::Impl::getLayerShapesRecursively Exception message: OpenCV(4.5.1) C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\layers\concat_layer.cpp:102: error: (-201:Incorrect size of input array) Inconsistent shape for ConcatLayer in function ‘cv::dnn::ConcatLayerImpl::getMemoryShapes’

Exception in thread “main” CvException [org.opencv.core.CvException: cv::Exception: OpenCV(4.5.1) C:\build\master_winpack-bindings-win64-vc14-static\opencv\modules\dnn\src\layers\concat_layer.cpp:102: error: (-201:Incorrect size of input array) Inconsistent shape for ConcatLayer in function ‘cv::dnn::ConcatLayerImpl::getMemoryShapes’
]
at org.opencv.dnn.Net.forward_4(Native Method)
at org.opencv.dnn.Net.forward(Net.java:289)
at com.jcarxmate.test.OpencvTest.test1(OpencvTest.java:105)
at com.jcarxmate.test.OpencvTest.main(OpencvTest.java:120)

However, with same logics, same opencv, same weights, same cfg, and same test video file in python. it works.

And if I test with yolov3-tiny in both Java and Python, everything works.
So, any suggestion?
Many thanks.