NVIDIA has created a neural network that turns a photo into a 3D model in seconds
The NVIDIA development team announced the launch of a neural network that can instantly turn any group of 2D photos into a 3D scene.
NVIDIA uses an inverse rendering technique to stack multiple photos taken from different angles and turn them into a fully-fledged 3D scene with correctly calculated lighting.
The company finalized the NeRF (Neural Radiance Field) machine learning model and presented its version of Instant NeRF at the GTC 2022 developer conference. The neural network spends only a few fractions of a second to analyze a group of static images and create their three-dimensional image. You can watch the presentation of Instant NeRF in this video . According to the company, this solution is today the fastest among similar ones.
At the Instant NeRF presentation, NVIDIA engineers also showed several scenarios for applying the new technology in real life. For example, you can create 3D avatars for VR applications or reconstruct real objects in 3D. The neural network can also be used to train robots and autonomous machines so that they can better recognize the size of real objects from photographs and videos.
