Time Does Tell: Self-Supervised Time-Tuning of Dense Image Representations
This addresses the gap in leveraging temporal data for dense representations, benefiting unsupervised segmentation and pretraining tasks, though it is incremental as it builds on image-pretrained models.
The paper tackles the problem of dense self-supervised learning by incorporating temporal consistency from videos, improving unsupervised semantic segmentation on videos by 8-10% and matching state-of-the-art on images.
Spatially dense self-supervised learning is a rapidly growing problem domain with promising applications for unsupervised segmentation and pretraining for dense downstream tasks. Despite the abundance of temporal data in the form of videos, this information-rich source has been largely overlooked. Our paper aims to address this gap by proposing a novel approach that incorporates temporal consistency in dense self-supervised learning. While methods designed solely for images face difficulties in achieving even the same performance on videos, our method improves not only the representation quality for videos-but also images. Our approach, which we call time-tuning, starts from image-pretrained models and fine-tunes them with a novel self-supervised temporal-alignment clustering loss on unlabeled videos. This effectively facilitates the transfer of high-level information from videos to image representations. Time-tuning improves the state-of-the-art by 8-10% for unsupervised semantic segmentation on videos and matches it for images. We believe this method paves the way for further self-supervised scaling by leveraging the abundant availability of videos. The implementation can be found here : https://github.com/SMSD75/Timetuning