I'm currently working on training a YOLOv8 model for my project. The goal is to train this model to recognize the product's stock code from a photo sent to a chatbot.
I have 1,522 stock codes, with about 10-15 photos for each stock code. Although the number of photos is a bit low, our customers typically send us the same photos used in training. For example, they might take a screenshot from our Instagram profile and send it to us.
I'm training the model, but it's only achieving around 70% accuracy. I believe the issue might be related to the hyperparameters, but I'm not very familiar with them. I would really appreciate any suggestions... (read 256 more words →)