Prompt: A potato factory has hired you to install a vision system for quality control and sorting. The potatoes are either red or gold $($yellow$).$ The potatoes move along a light grey conveyer belt. The potatoes need to be sorted for color, size, and shape. While there are many potatoes on the conveyor belt at once, the potatoes are not touching, and each potato is completely visible to the camera $($no occlusions$).$

Potato : Could you use the PCA potato embedding space to estimate potato quality? Assume you have training images of potatoes of both types $($red and gold$),$ ranked from $1($low$)$ to $5($high$)$ in terms of quality. Would you use a single embedding space, two embedding spaces, multiple embedding spaces? Explain your setup and process.