The images generated by provided pre-train checkpoints are not comparable to the samples shown in the paper. Following are the results for 64x64 and 256x256 respectively. And the script for testing is python test.py --dataroot /project/dzhang4/vast_data/dayton_5 --name dayton_g2a_256_pretrained --model selectiongan --which_model_netG unet_256 --which_direction BtoA --dataset_mode aligned --norm batch --gpu_ids 0 --batchSize 4 --loadSize 286 --fineSize 256 --no_flip --eval
The images generated by provided pre-train checkpoints are not comparable to the samples shown in the paper. Following are the results for 64x64 and 256x256 respectively. And the script for testing is
python test.py --dataroot /project/dzhang4/vast_data/dayton_5 --name dayton_g2a_256_pretrained --model selectiongan --which_model_netG unet_256 --which_direction BtoA --dataset_mode aligned --norm batch --gpu_ids 0 --batchSize 4 --loadSize 286 --fineSize 256 --no_flip --eval