Bin Li

Real-time, lightweight image compression is now possible with diffusion models, thanks to a novel architecture that swaps transformers for convolutions and prioritizes compression-focused pre-training.

Event cameras can now accurately measure high-speed 3D deformations of structures under extreme lighting, opening up new possibilities for monitoring the safety of critical infrastructure.

Achieve higher accuracy in dual-arm robot calibration by unifying coordinate and kinematic parameter estimation within a single Lie-algebraic framework, eliminating artificial error separation.

Ditching the 2D latent grid unlocks 60%+ bitrate reductions in generative video compression by encoding videos into adaptable 1D latent tokens.

Forget trying to wrangle dynamic 4D scenes with recurrent networks – DynamicVGGT achieves state-of-the-art reconstruction accuracy using a surprisingly effective feed-forward approach.

Stop naively aggregating knowledge for KB-VQA: MaS-VQA's Mask-and-Select mechanism shows how to prune irrelevant image regions and knowledge fragments for better reasoning.

MLLMs can now spot subtle image forgeries with SOTA accuracy by strategically using forensic tools to expose hidden inconsistencies, outperforming traditional text-centric approaches.