anonseoul/turn-taking-dataset

turn-taking-detection-dataset scripts/notebooks for turn taking detection dataset Frame Feature extraction Currently, extracting feature with mmaciton: current method needs to be fix. It is difference from the OAD convention.current method: please refer thisOnly taking the central frame among non-overlapping video snippets. In OAD task, all the frames in from snippets are converted into the mean of features from them. To-dos (legacy, now maintain @… See the full description on the dataset page: https://huggingface.co/datasets/anonseoul/turn-taking-dataset.