After training when I plot the top losses, all the images had this weird pinkish texture. Are the images being automatically augmented in some way that I don’t want?
This also results in the validation loss being too high, in the 70’s
This also results in the validation loss being too high, in the 70’s
That’s your heat map, it shows what your model is looking at. When you call plot, I believe you can pass in heatmap=False