#ProjectCloak
summary
Project Cloak is a research prototype that does content-aware fill for video, taking two inputs: source video and mask and producing temporally consistent frames that remove objects instead of patching each frame independently. It needs a tracked mask (polygon mask/rotobrush) and uses learned priors to imagine what was behind the occluder, handling straps, stains, shadows, tripod/boom removal and even full-person removal, showing how a single-pass algorithm can replace tedious per-frame retouching.
AI-written summary · about
chapters (4)
related talks
28mBackground removal without background knowledge
40mSmart Crop and Smart Preview via Video Understanding
25mOptimal Video Compression Using Pixel Shift Tracking
21mPmomo: Projection Mapping on Movable 3D Object
18mCoVA: Exploiting Compressed-Domain Analysis to Accelerate Video Analytics
58mPushing Deepfakes to the Limit-Fake video calls with AI
28mFace spoofing detection: theory and practice
33mArchitecting Scalable Multi-Modal Video Search