OpenAI announces Point-E, a machine learning system that rapidly generates 3D images from text prompts

OpenAI announces Point-E, a machine learning system that creates 3D images quickly with text prompts

High-level overview of pipelines. First, a text prompt is injected into the GLIDE model to generate the composite rendering view. Next, a conditional point cloud diffuses stack over this image to create a 3D RGB point cloud. Credit: arXiv (2022). DOI: 10.48550/arxiv.22212.08751

A team of researchers at San Francisco-based OpenAI, has announced the development of a machine learning system that can generate 3D images from text much faster than other systems. The team published a paper describing their new system, called Point-E, on arXiv print server available.

Over the past year, several groups have announced products or systems that can generate 3D model images based on text prompts, for example, “blue chair on red floor” or “boy wearing a green hat and riding a purple bicycle.” Such systems usually have two parts. The first person reads the text and tries to understand it. Second, get trained in internet search, show the desired image.

Due to the complexity of the task, these systems can take a long time to return a model, from hours to days. In this new effort, the researchers built a similar system that returns results within minutes, although they readily admit that the results “lack modern technology in terms of sample quality.”

To generate images faster, the researchers adopted a slightly different approach than others. Their system doesn’t even produce images in the traditional sense. Instead, it creates point clouds that, when viewed together, resemble the desired image. The team took this approach because it’s much easier to create point clouds than it is to create actual images. To produce the results, the system routes the images it finds through another AI system they have developed that converts what it receives into meshes, creating a point cloud model. 3D of the desired object.

The first part of the system is created using two modules—the first part converts text into visual ideas, and the second finds images used to create a generic image. In operation, the system runs very much like other systems of its kind—users enter a descriptive text prompt and the system returns a visual model. They note that while image quality cannot be compared with other systems, it may be more suitable for other applications, such as fabricating real-world objects through a 3D printer.

The researchers created the system open access—users who want to work with it can access the above code GitHub.

More information:
Alex Nichol et al, Point-E: System for generating 3D point clouds from complex prompts, arXiv (2022). DOI: 10.48550/arxiv.22212.08751

Journal information:

© 2022 Science X Network

quote: OpenAI announces Point-E, a machine learning system that rapidly generates 3D images from text prompts (2022, December 21) retrieved December 21, 2022 from /2022-12-openai-point-e -machine-fast-3d.html

This document is the subject for the collection of authors. Other than any fair dealing for private learning or research purposes, no part may be reproduced without written permission. The content provided is for informational purposes only.


News7F: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button