Alexander Alekhin

You could find some details here: https://github.com/openvinotoolkit/cvat/issues/4444#issuecomment-1141066544

BTW, libjpeg-turbo 2.1.4 has been released: https://github.com/libjpeg-turbo/libjpeg-turbo/releases/tag/2.1.4

Thank you for the report! BTW, It makes sense to check (and report to) the upstream project: https://github.com/uclouvain/openjpeg

BTW, There is Darknet importer in OpenCV (and tests for YOLOv4). > converted yolov4 (custom trained on our dataset) from darknet to onnx

FOURCC is a messy area. It would be nice if we could clarify that. There is https://www.fourcc.org/codecs.php but it doesn't know about VP9. FFmpeg has own entries for that. AFAIK,...

I believe we could start from tests with parameters to see the scope of the problem: - `std::tuple` Test files for different codecs can be generated through ffmpeg (use `generateFrame()`...

You could check this PR (and its tests): #4089 P.S. On Windows OpenH264 binaries (.dll files) are required to enable H264 encoding. Version is defined here: https://github.com/opencv/opencv_3rdparty/blob/ffmpeg/4.x/ffmpeg/download_src.sh#L26

Feel free to prepare PR with proposed changes to "4.x" branch.

@eliteraspberries Don't worry. It is more easy to discuss changes through PR. Problem of current conditions is that they don't expect case then `yStride != width` (1472 vs 1440). However,...

There is some problem with script: ``` Model expect fcd0fa401c83bf2b29e18239a9c2c989c9b8669d catch [Errno 2] No such file or directory: 'onnx/models/yolov7_not_simplified.onnx' hash check failed - downloading get https://drive.google.com/u/0/uc?id=1rm3mIqjJNu0xPTCjMKnXccspazV1B2zv&export=download 200 OK [ Mb]...