What I would like to achieve is this:

  1. a single camera feeds a live video stream to the computer
  2. the geometry of the scene is determined with some free photogrammetry library
  3. the objects are recognized with YOLO
  4. the scene and objects are reconstructed with simple shapes such as elongated spheres, pyramids, cones, cylinders and cubes

I am a newbie in the field of robotics. I am not sure if this is the right way to do things. I would appreciate any advice and suggestions. Particularly, I would be really grateful if someone could point me to the right up-to-date library to use. The library can be in Python, C++, C#, Java, JavaScript and so on.