Sign In

Communications of the ACM

ACM TechNews

Natural World Empowers Future of Computer Vision

View as: Print Mobile App Share:
The new open-source software system generates an infinite number of photorealistic scenes of the natural world.

The images generated by the system and automatically labeled can then be used to train a robot to recognize and locate objects using only an image as input.

Credit: Jia Deng et al

An open source software system developed by Princeton University researchers aims to improve the training of computer vision systems by quickly generating an unlimited number of photorealistic scenes of the natural world.

The Infinigen system uses randomized mathematical rules to produce natural-looking three-dimensional (3D) objects and synthetic representations of natural phenomena.

Each image is generated programmatically, which involves developing a virtual world and populating it with digital objects, which allows for automatic labeling and training computer vision systems to identify and locate objects with only an image as input.

Princeton's Jia Deng said the researchers expect Infinigen will prove “a useful resource not just for creating training data for computer vision, but also for augmented and virtual reality, game development, film-making, 3D printing, and content generation in general."

From Princeton University
View Full Article


Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account