New AI Model for Human Reconstruction in 3D

"Recent advances in image-based 3D human shape estimation have been driven by the significant improvement in representation power afforded by deep neural networks," wrote the team. "Although current approaches have demonstrated the potential in real-world settings, they still fail to produce reconstructions with the level of detail often present in the input images" 

The research team noted that the main problem here involves two conflicting requirements. Accurate predictions require a large context, but at the same time, precise predictions require high resolution. Current hardware has memory limits, so we need a new model. 

The new paper proposes using a multi-level architecture that is end-to-end trainable. First, there's a coarse level that studies the whole image at a lower resolution and focuses on holistic reasoning to provide "context to a fine level which estimates highly detailed geometry by observing higher-resolution images."

They state that the new approach outperforms current "state-of-the-art techniques on single image human shape reconstruction by fully leveraging 1k-resolution input images."

You can find the full paper here. Don't forget to join our new Telegram channel, our Discord, follow us on Instagram and Twitter, where we are sharing breakdowns, latest news, awesome artworks, and more.