Kagami
That's probably due to RGB -> YCbCr 4:2:0 conversion step. I need to implement loading Y4M source files, that way you should have bit exact Y4M after decoding AVIF.
1st and 3rd cases seem to be the same? Also, I don't particularly like that the whole loop is duplicated... For best performance both decoding and colorformat conversion should be...
I think that would be good addition to go-face. These features are often useful in face recognition. I don't have much time to work on this unfortunately. But would be...
Landmarks (face shapes) are supported.
Yeah, good idea, we may want to return few most probable matches.
Need to rewrite this: ```cpp std::string dir = model_dir; std::string shape_predictor_path = dir + "/shape_predictor_5_face_landmarks.dat"; std::string resnet_path = dir + "/dlib_face_recognition_resnet_model_v1.dat"; std::string cnn_resnet_path = dir + "/mmod_human_face_detector.dat"; deserialize(shape_predictor_path) >> sp_;...
Hi. I haven't heavily tested multiple recording sessions on the same page, there might be some bugs. I will look into it, thanks.
Technically yes, but currently there is no API for that. You could look at implementation of `Recorder` class. It handles `onaudioprocess` event of its internal audio node and might be...
Yes, it's possible, just need to make `vmsg_encode` C function return the number of bytes written, so you can send `v->mp3+v->size-n .. v->mp3+v->size` bytes via PostMessage to the main thread....