Seon Joo Kim

G-NoCL uses generative models to enhance continual learning without extensive labeled data.

Using unlabeled videos to improve action recognition in lengthy videos.

ActionSwitch detects actions in streaming videos without needing prior class information.

New method improves visibility of hidden objects using lasers and advanced algorithms.

A new method revolutionizes 3D modeling using casual video input.

Revolutionary model creates realistic talking head videos at high speed.