There are measuring machines which, based on sensors, are able to determine specific parameters of the body shape or state.
(proteins, muscle mass, adipose tissue mass, body-weight, height)
and separate into segments. (right hand, left hand etc.)
Hovewer the same issue seems relatively difficult.
With my humble opinion, I can recommend the process to you:
- Prepare a position pattern - all photos should be in a similar
pattern based on one algorithm.
- Reading height or weight itself seems to be a complicated topic.
- I suggest Initially start on the algorithm - image recognition(You can reuse existing solutions (Pytorch || TensorFlow) - (CNN + seq2seq with visual attention will be usefull here)
- Then use some methods - which can be seen in the photo - after obtaining (x1, x2, y1, y2) from the person - measure the correlation between y and y axis (the greater the higher the weight - although not necessarily)
While exploring, more variables will appear for analysis
Initially recommend 2 repositories: