I'm using transfer learning with VGG16 for image classification. I have 6 classes each one with more than 20k images, I'm trying to improve my accuracy but after many tests I still don't have good results. My accuracy is around 60% or sometimes less than that. I did many tests on the learning rate, batch size, epochs. In fact, I want to classify parts of cars by View ( front, back, left,...) so for example in class front I have different parts of car (door, window, trim ...) where the picture was token on the front side. Maybe this doesn't help the model to learn well because there are images similar for example in left and right classes. I hope that someone could help me to improve it.

  • $\begingroup$ There are a lot of specifics regarding your question and its too general to give a proper answer. There is a really nice article regarding same problem with code and explanation, maybe it can give you a hint statworx.com/en/blog/… $\endgroup$ May 18, 2021 at 8:55
  • $\begingroup$ ok, thank you. I will see it. $\endgroup$
    – Lema Zaidi
    May 18, 2021 at 11:29


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.