NVIDIA’s AI Revolution

Get ready for a revolution in visual AI!

NVIDIA is pushing the boundaries at CVPR with groundbreaking research across image generation, 3D scene editing, and autonomous vehicles.
“Generative AI is a game-changer,” says Jan Kautz, VP of NVIDIA’s AI research. “At CVPR, we’re showcasing how we’re empowering creators and building the future of self-driving cars.”

Here’s a glimpse into NVIDIA’s arsenal of over 50 research projects:

• Customizable Image Creation: JeDi lets creators personalize text-to-image models – think generating specific objects or characters – in record time, ditching complex fine-tuning processes.
• Instant 3D Pose Estimation: FoundationPose instantly grasps and tracks object poses in videos, no per-object training needed. This paves the way for groundbreaking AR and robotics applications.
• Effortless 3D Scene Editing: NeRFDeformer edits 3D scenes captured by Neural Radiance Fields (NeRFs) with a single 2D image. No more manual reanimation – just a snap and your scene is transformed!
• Vision & Language United: VILA, co-developed with MIT, is a family of vision language models that masters understanding images, videos, and text. It can even decipher internet memes – talk about intelligence!

NVIDIA’s research extends beyond captivating visuals. Over a dozen papers explore innovative solutions for autonomous vehicle perception, mapping, and planning. Sanja Fidler, VP of NVIDIA’s AI Research, highlights the potential of vision language models for self-driving cars.

The breadth of NVIDIA’s work at CVPR showcases the immense power of generative AI. From empowering creators to accelerating automation and propelling self-driving cars, NVIDIA is shaping the future of visual intelligence.