This project was originally a Meta research paper published in May 2020 as PHOSA . The basic idea is to convert 2d images to 3d meshes for images that includes humans interacting with specific objects like baseball bat, bicycle, tennis racket, motorcycle and others. After understanding the most of the code I added the app section using Flask I also increased the number of objects that the model can accept to 8 instead of 1( bicycle ). I've accomplished this project on AWS EC2 computing unit-linux based system.
github repo: https://github.com/MohsenAziz/2d-imag...
colab notebook: https://colab.research.google.com/dri...