NVIDIA is showcasing its latest advancements in visual artificial intelligence at the Computer Vision and Pattern Recognition conference in Seattle, featuring new models and techniques in areas such as custom image generation, 3D scene editing, and autonomous vehicle perception. The company's researchers are presenting over 50 projects, including two finalists for Best Paper Awards and a winning entry in the Autonomous Grand Challenge. Notable breakthroughs include a technique to rapidly customize diffusion models and a foundation model that can instantly understand and track 3D object poses in videos.