Facebook has launched a new technology which creates 3D models of human photos, dubbed PIFuHD, that works through multi-level architecture based on deep neural network to allow users create fully-detailed 3D models of a photograph.
While the PIFuHD technology which stands for Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization will be useful in automated image digitization in fields like medical imaging and virtual reality.
The 3D image conversion demo has been made available on Google Colab for users to try out the new technology, but as it's still in initial phase, there are few limitations, for instance, the technology currently can't accept images of over 1K resolutions.
What's the underlying technology behind PIFuHD?
Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization (PIFuHD) involves end-to-end trainable refined framework for implicit surface learning for high-resolution 3D human photo reconstruction at 1K image resolution.
The technology is perhaps one of the most revolutionary achievements of 2020 as it is capable of changing the way virtual reality is perceived and work in the areas of medical imaging.
Though, there are still some limitations in the new technology, as currently, it can't accept images with resolutions greater than 1K and it first translates the image into the front and back perspective to deliver the 3D figure in high resolution.
What's the Future projections for PIFuHD?
The major breakthrough behind this technology is the accurate predictions based on context, albeit the memory limitations in current hardware is a huge challenge presently.
PIFuHD technology, however will get to an advanced level, whereby changes will be made in the hardware and such memory limitations will be a thing of the past. If you wish to give the new technology a spin, you can get more information about it from this Facebook research resource.