research Loop Consistent Video Generative Model Joint training of video and multi-view native Diffusion Forcing Transformer for enabling complex loop closures via context enhancement. Self-Supervised 3D Occupancy Prediction Self-supervised 3D occupancy prediction for self-driving using SRT-based volumetric rendering. Speeding Up Adversarial Attacks 20x speedup of adversarial attacks against CNNs via advanced initialization while matching SOTA accuracy. Self-Supervised Peak Prediction for Mass Spectrometry Self-supervised GPT2-based model for predicting missing peaks in mass spectrometry data. industry Generative Furniture Removal Image editing pipeline for removing furniture from real estate photos. Shipped to 220M users. Multi-View Consistent Image Editing Multi-view-consistent image editing. 3D Scene Understanding with VLMs Agentic system for chatbot interaction with 3D scenes via 3D-grounded open-vocabulary detection with VLMs. fun Smart Gym Interactive workout system combining computer vision with a Unity game for real-time exercise tracking and gamification. Vánoční vzpomínka Advised my wife with her vibing this cute Christmas generative video page.
