Aiming to bridge the gap between human and machine vision in the longstanding problem of 3D human face and body reconstruction from 2D imagery, the project’s goal is to build an environment for benchmarking reconstruction methods.
The project proposes the development of a software toolkit and web interface able to properly evaluate the fidelity of a 3D human face and body reconstruction, from both the modelling of the anthropometric dimensions and the preservation of the phenotypic traits (identity). The motivation comes from the fact that such methods are usually non-sensitive to face identities and strongly impacted by facial variations, but also when the reconstruction task is targeted for the whole human body, it introduces a particular group of challenges due to the skeleton structure.
The requirements may include:
- Defining the API (based on photogrammetric decomposition of texture maps / UV maps) (R&D);
- Defining the comparison methodology (R&D);
- Developing the libraries for handling 3D models (OBJ Files and MTL File Format), 3D to 2D projections (SW);
- Implementing the comparison methodology (SW);
- Setting up an evaluation server (Dev Ops);
- Developing a basic front end in order to access the server through via web (Dev Ops);
Skills required / What you will learn:
- C++ / Python Programming (mid-level / advanced)
- Web Development / Networking (entry-level)