NVIDIA announces Neuralangelo AI model that can convert video content into high-precision 3D models

IT House News on June 2, Nvidia announced this week the launch of an AI model called Neuralangelo, which can convert video clips into 3D models with rich levels of detail and textures.

▲ Picture source NVIDIA

IT House inquired and learned that there have been several AIs that generate 3D models based on video content, but these AIs often cannot start when faced with low-contrast videos. NVIDIA's Neuralangelo model uses instant neural graphics primitives and utilizes its own Instant NeRF technology, so even low-contrast videos can capture subtle details and convert them into more commercial Value 3D model.

▲ Picture source NVIDIA

NVIDIA emphasizes that the uniqueness of Neuralangelo is mainly reflected in its "accuracy" and "universality":

• "Accuracy" means that the model can accurately represent the texture of complex materials, such as roof tiles, glass panes, and smooth marble;
• "Universal" means that creators can create realistic objects simply using video footage shot with smartphones and integrate them into their projects.

▲ Picture source NVIDIA

The Neuralangelo model uses videos of items taken from different angles as input material to generate the models that creators need. Once Neuralangelo determines the movement of the camera in the video, it generates a 3D outline of the model and begins carving out details on it.

▲ Picture source NVIDIA

Nvidia is expected to release the Neuralangelo model at the Computer Vision and Pattern Recognition (CVPR) conference held from June 18 to 22. In internal demonstrations, NVIDIA showed 3D restored models of everyday items such as Michelangelo's famous sculpture "David" reconstructed by the Neuralangelo model, NVIDIA's offices in the Bay Area, and flatbed trailers.

▲ Picture source NVIDIA

In addition to Neuralangelo, the models displayed by Nvidia at the conference also included one called DiffCollage. DiffCollage is an AI model that can generate videos through text prompts. It can generate 360-degree panoramic images and looping dynamic images, and then expand these images into specific videos.

Ming-Yu Liu, senior director of NVIDIA Research and author of the paper, gave his thoughts on the significance of this research:

The 3D reconstruction capabilities provided by Neuralangelo will bring huge benefits to creators, helping them recreate the real world in the digital world. This tool will eventually allow developers to import detailed objects – whether they are small statues or large buildings – into virtual environments in video games or industrial digital twins.

IT House friends can find the specific content of the paper here.

The above is the detailed content of NVIDIA announces Neuralangelo AI model that can convert video content into high-precision 3D models. For more information, please follow other related articles on the PHP Chinese website!