Advices needed to improve model

I recently train a classifier (with the dataset that I scrapped myself) using some pretrained models. However the accuracy did not able to break through 60% while the losses showed that the model is overfitting… I trained, adjusted lr using lr_find, then trained again, but the metrics are still not improving. Is it because of the problem on the datasets or any other possible reason?
Appreciate a lot if you are kind to point me some direction to move on.

Try to gather more data and/or reduce the model complexity.
This might help:

1 Like