Saliency map seems not improving with more epoch

Hello all,
I am working on the coding assignment of the saliency maps. My do_saliency() function seems generating acceptable images with the weight from 0_epoch. However, when I load weights from 15_epoch or 95_epoch, the images seems worse. Please find my saliency map at 0, 15, 95 epoch below.
Could you please kindly let me know what might be wrong here? I don’t know where to debug at this point. Thanks a lot.

Please check the validation loss and accuracy when you train your model for larger number of epochs.
One possible reason for poor saliency maps could be that your model is overfitting on the training data and unable to generalize well.

1 Like

I tested the trained model with validation set. my accuracy is more than 0.80 and the loss is also reasonably close to the loss of training set. Does this mean something wrong with my do_saliency() function?

If your model performance on the training set is close to the performance on the validation set, then you need to check the implementation of the do_saliency function.