when AI is used like google assistant is used it records our human voice it learns our human voice and when an incoming call is received and ask to reply to that call it uses the same human voice recorded by the user in that particular device. my question is that when an AI can use a human voice for replying a call then why can't it use the face recorded during the face recognition in a particular device for the virtual actor
Yes, after a fashion. It is possible to take a 2d image of a face and create a 3d model of the entire head. SHBMM (Spherical Harmonic Basis Morphological Models) are like 3dFFTs done with a linear combination of parameterized spheres instead of a linear combination of sine waves. This is currently used primarily for generating head models to address lighting effects. Many papers are available by googling SHBMM. It's not a stretch to see such techniques used to generate real time models. Take a lot of computation to do in real time, but that's what GPUs are for!
Also deep learning can be involved in turning of 2d image to 3d model (https://github.com/AaronJackson/vrn , https://github.com/dezmoanded/vrn-torch-to-keras) or even for more direct transformation of one image to another image (see for "deep fake")
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning