2.5D-Visual-Sound
2.5D-Visual-Sound copied to clipboard
The AudioVisualDataset class tries to open non-existing frames
Hello,
I haven't digged too much yet, but it seems that the AudioVisualDataset.getitem method opens an image (frame) that doesn't exist:
frame = process_image(Image.open(os.path.join(frame_path, str(frame_index).zfill(6) + '.png')).convert('RGB'), self.opt.enable_data_augmentation)
I didn't find any code the creates these images. Am I supposed to implement logic to create them?
From what I've experienced, you have to use something like ffmpeg to create them yourself