Nextech3D.AI (OTCQX: NEXCF) (CSE: NTAR) (FSE: EP2) filed a patent for using artificial intelligence to create complex 3D models from 2D images for use in ecommerce.
Filed on Thursday, the provisional patent is called “Generative AI for 3D Model Creation from 2D Photos using Stable Diffusion with Deformable Template Conditioning.”
The patent is for a solution for creating 3D models from 2D reference photos, even as a whole or in parts, by revolving deformable templates to convert into three dimensions. It uses recent advances in generative AI and in particular, diffusion models to accomplish this.
“We have been preparing for 5-years, anticipating this exact moment in time, where our AI technology would advance far enough to be patentable while simultaneously seeing market demand for photorealistic 3D models for ecommerce pick-up momentum,” said Evan Gappelberg, Nextech3D.ai CEO.
Read more: Powerbridge Technologies integrates ChatGPT for online education services
Nextech builds a library of photorealistic 3D assets
Originally called Nextech AR Solutions, Nextech3D.ai uses proprietary AI to create 3D experiences for the metaverse.
Machine learning employs diffusion models to analyze datasets and uncover patterns in the data.
Reference photos of an asset in the library. Image via Nextech3D.ai.
Nextech has built tens of thousands of high quality, photorealistic 3D assets, with hundreds of thousands of individual parts. Further, the company is putting them into a library, and synthetically rendering them from random views. Afterwards, the company will use the data to train new diffusion models capable of reconstructing 3D mesh parts from reference photos.
Collection of all unique parts of a 3D digital twin. Image via Nextech3D.ai.
The company’s first clean dataset has over 70,000 3D objects and over 2.2 million rendered reference photos ready for training.
“Today, we now see both the demand side of the 3D model market and the supply side of 3D model production fuelled by ROI and AI, lining up for a breakout year in revenue and cash flow for Nextech3D.ai,” said Gappelberg.
Each part is rendered from random views. Image via Nextech3D.ai
Additionally, the company is working on other AI models to prescribe functions to assemble parts into full 3D models.
Follow Joseph Morton on Twitter
joseph@mugglehead.com
