I am doing a project to improve the resolution of images clicked by my camera phone. Most of these images are 12MP in size, but to fit them on the GPU, I am resizing them to size 512 (for learner 1a & 1b) and size 1024 (for learner 2a & 2b).
My convergence on 2a is poor with default parameters so I decided to run a lr_find() on the pretrained 1b modeland this is what I got:
Please suggest what LR should I pick?
