Labelling unlabelled videos from scratch with multi-modal self-supervision
This addresses the high cost of human annotation for videos, offering a method for pseudo-labeling without supervision, though it is incremental as it extends image-based techniques to the video domain.
The authors tackled unsupervised labeling of video datasets by proposing a novel clustering method that leverages audio-visual correspondence, achieving high semantic overlap with ground truth labels on benchmarks like Kinetics and VGG-Sound.
A large part of the current success of deep learning lies in the effectiveness of data -- more precisely: labelled data. Yet, labelling a dataset with human annotation continues to carry high costs, especially for videos. While in the image domain, recent methods have allowed to generate meaningful (pseudo-) labels for unlabelled datasets without supervision, this development is missing for the video domain where learning feature representations is the current focus. In this work, we a) show that unsupervised labelling of a video dataset does not come for free from strong feature encoders and b) propose a novel clustering method that allows pseudo-labelling of a video dataset without any human annotations, by leveraging the natural correspondence between the audio and visual modalities. An extensive analysis shows that the resulting clusters have high semantic overlap to ground truth human labels. We further introduce the first benchmarking results on unsupervised labelling of common video datasets Kinetics, Kinetics-Sound, VGG-Sound and AVE.