Hello, I work on AI systems and develop models for years, and I have already done image classification tasks. You can check my github page: unnamed_idea. As an example, I develop a model for traffic sign classification, which takes an image and categorizes it among 40+ traffic signs.
Thanks to my background knowledge, I already have written image augmentation functions, thus your input will be around 8 times larger than what you deliver to me:) As a result the model will be robust and extremely precise.
As you have said, you can send me some of the images and I can share you initial model / results in a short notice.
Again, you can check my github page, we even develop self driving car models based on cars' cameras. Waiting for your answer and some initial data,
best regards:)