Facebook has developed a new technology that can create 3D models of human beings from their photographs. According to the reports by Slashdot, the new technology is dubbed as PIFUHD that is based on a deep neural network that allows it to create a highly detailed model from a photograph.
This technology can be used in the field of virtual reality and medical imaging and a detailed report of the research has been released with the GitHub link.
Working of PIFuHD
PIFUHD stands for Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization. The technology creates a first and back perspective of the image and delivers a 3D model image as output.
The resultant 3D image created by this tool was able to create a detailed 3D model and was sufficient enough to get detailed information such as fingers, facial features, and clothing folds.
Limitation of the PIFUHD Innovation
The resolution of the image should be between 512×512 and 128 × 128 in resolution respectively due to the memory limitation of the graphics hardware.
To run the PIFUHD model at least 8GB GPU memory is recommended. A pre-trained model is also available on the GitHub link.
A demo version of the model is also available on Google Colab for other users to try.
