# First training on augmented set
CUDA_VISIBLE_DEVICES=0,1,2,3 python train.py --dataset pascal_aug --model fcn --aux --backbone resnet50 --lr 0.001 --checkname mycheckpoint
# Finetuning on original set
CUDA_VISIBLE_DEVICES=0,1,2,3 python train.py --dataset pascal_voc --model fcn --aux --backbone resnet50 --lr 0.0001 --checkname mycheckpoint --resume runs/pascal_aug/fcn/mycheckpoint/checkpoint.params