San Francisco

A non-profit, artificial intelligence (AI) research company, OpenAI which is also behind creating text-to-image generator, DALL-E and AI-powered ChatGPT chatbot has announced major changes in its 3D rendering which has led to Point-E. The machine learning system creates a 3D object based on a text prompt. 

Advertisment

According to a paper, published by the company, Point-E uses just one single Nvidia V100 GPU to process images and complete generating them in minutes which would typically take hours and require multiple GPUs. While there are not actually 3D models as they do not capture the texture and detail of what is entered it does so by using Point Clouds, which are sets of data points in space that represent a 3D shape. 

ALSO READ | Explained: What is Lensa AI and should we be concerned about it?

The upside of using Point Clouds is that they require significantly less computational power, since the name Point-E where the ‘E’ stands for “efficiency”. Therefore, while the accuracy is a key limitation of the AI-powered tool, the speed is note worthy. 

Advertisment

However, in order to overcome this the team trained two AI models to convert the point clouds into 3D meshes. According to TechCrunch, these Meshes are, “the collections of vertices, edges and faces that define an object – are commonly used in 3D modelling and design.” This might lead to some bulky or distorted shapes, the paper noted. 

ALSO WATCH | WION Fineprint | AI that can detect gun violence?

The two AI model utilises a text-to-image and image-to-3D approach after which the first tool creates a rendered object that best fits the data entered. Subsequently, the image-to-3D model, which is already programmed with a set of images paired with 3D objects learns to effectively translate between the two, creating a Point Cloud based on them and converting it into a Mesh.

Advertisment

“While our method still falls short of the state-of-the-art in terms of sample quality, it is one to two orders of magnitude faster to sample from, offering a practical trade-off for some use cases,” said the paper. It added, “This could make it more practical for certain applications, or could allow for the discovery of the higher-quality 3D object.” 

ALSO READ | What is ChatGPT, an AI chatbot that has taken internet by storm? Can it replace humans?

While OpenAI’s Point-E is the latest to opt for the growing trend of AI-generated 3D objects it certainly isn’t the only one. Earlier this year, Alphabet Inc’s Google released Dream Fusion, which was an expanded version of its predecessor Dream Fields released a year prior. 

However, unlike Dream Fields, DreamFusion could generate 3D representations of objects without 3D data. Reports suggest that the efficiency of Point-E beats Dream Fusions which requires a lot of time as well as GPUs to produce the images. 

(With inputs from agencies) 

WATCH WION LIVE HERE

You can now write for wionews.com and be a part of the community. Share your stories and opinions with us here.