Closed khawar-islam closed 10 months ago
Dear @3neutronstar When I trained resnet50 (Vanilla) model from scratch on Stanford Car dataset, the model achieved 80.92% results and in your paper resnet50 (Vanilla) accuracy is 85.52% which is quite different. That is my problem I am facing right now
Dear @khawar-islam,
Thank you for your suggestion. I will upload FGVC implementation soon.
Dear @3neutronstar
thank you very much, I will be waiting for your code
Dear @3neutronstar
Would it be possible for you to give me a favor? As I have searched a lot on GitHub regarding FGVC code and mostly codes are utilizing pre-trained model to fine-tune on FGVC dataset. When i trained from scratch the accuracy significantly dropped, code you please your FGVC codebase?
Regards, Khawar