Do we still need ImageNet pre-training in remote sensing scene classification?

Re-assign community