HOW TO stitch enhanced patches to single image?

aiff22 / DPED

Software and pre-trained models for automatic photo quality enhancement using Deep Convolutional Networks

1.68k stars 369 forks source link

HOW TO stitch enhanced patches to single image? #20

Closed kirill-pinigin closed 5 years ago

kirill-pinigin commented 5 years ago

Please, tell me HOW TO stitch enhanced patches to single image? What do you use to stitch enhanced patches to single image? Do you meet some problem with combining or stitching? Or May be some artifacts?

aiff22 commented 5 years ago

@kirill-pinigin, you should pass the entire (full-resolution) image to the network instead of stitching the patches. Please note that the latter option will result in a corrupted image: instance normalization layers used in DPED network are leading to different brightness of the patches, and thus after combining them you will get an image with a checkerboard pattern.

kirill-pinigin commented 5 years ago

@aiff22 will it work if i replace instance_norm with batch_norm and retrain generator via your framework? Can I eliminate checkboard artifacts?