Hi, we are building an image detection + recognition system for a grocery store. We need to identify $10 Lays Packet and $20 Lays Packet of the same flavour as different classes. Considering the fact that these both products would have same packaging, do you think they can be distinguished as separate classes? The only difference would be size of the packet and the weight. We feel that while running inferences in real time, the bounding box dimensions would change continuously, so relying on bounding box area would not be a feasible way to go about it. What would be a good solution to this problem?
the solution would be a neural network and a lot of training and since this is a grocery store, we are moving towards tensorflow and cluster computing with X nodes to accelerate learning