NVIDIA research leads in visual generative AI, showcasing projects at CVPR conference.

From NVIDIA: 2024-06-17 09:00:36

NVIDIA researchers are leading in visual generative AI, showcasing more than 50 projects at the CVPR conference. Papers on diffusion models and HD maps for autonomous vehicles are finalists for Best Paper Awards, with NVIDIA winning the Autonomous Grand Challenge. Their work includes text-to-image models, object pose estimation, NeRF editing, and visual language models with industry-specific applications.

JeDi simplifies custom image generation by allowing users to personalize diffusion models with reference images in seconds. FoundationPose is a new model for object pose estimation and tracking that doesn’t require fine-tuning and sets a new benchmark record. NeRFDeformer simplifies transforming 3D scenes with a single snapshot, while VILA visual language models outperform prior networks in answering image questions.

NVIDIA contributes to autonomous driving research with a multitude of papers and developments, including the largest indoor synthetic dataset for the AI City Challenge. They also introduce NVIDIA Omniverse Cloud Sensor RTX microservices for sensor simulation, advancing the development of autonomous machines. NVIDIA Research continues to push the boundaries in AI, computer vision, self-driving cars, and robotics.



Read more at NVIDIA: NVIDIA Research Showcases Visual Generative AI at CVPR