On Monday, 29 June 2026, at 11:00 AM, the Center for Computational & Data Sciences (CCDS), Independent University, Bangladesh (IUB), hosted a session on self-supervised visual learning from video motion, presented by Md Raqibul Islam.
Presenter: Md Raqibul Islam, Research Assistant, CCDS, IUB
Paper Presented: You Don't Need Strong Assumptions: Visual Representation Learning via Temporal Differences
The session featured a discussion on Temporal Difference in Vision (TDV), a self-supervised learning paradigm that learns visual representations directly from natural motion in video, without relying on hand-crafted augmentations, masking, or cropping. Discussion covered the paper's central argument that weaker inductive biases scale better with larger datasets, the architecture's jointly trained frame and motion encoders, and its competitive performance on dense spatial tasks such as segmentation, optical flow, and stereo depth.
We thank all participants for their valuable contributions and insightful discussions.
Tags: CCDS Weekly Paper Presentation, Artificial Intelligence, Machine Learning, Self-Supervised Learning, Computer Vision, Research, Deep Learning, CCDS, IUB