Hi Doubiiu,great work!
I find it that in scripts/evaluation/inference.py ,we don't transform the condition image to [-1,1] (such as frame_tensor = (frame_tensor / 255. - 0.5) * 2).However,in scripts/evaluation/funcs.py , the function load_image_batch applies a transformation at line 201 , img_tensor = (img_tensor / 255. - 0.5) * 2 . I wonder whether or not there is a need to apply this at the inference code.
Many thanks for you time!
Hi Doubiiu,great work! I find it that in
scripts/evaluation/inference.py
,we don't transform the condition image to [-1,1] (such asframe_tensor = (frame_tensor / 255. - 0.5) * 2
).However,inscripts/evaluation/funcs.py
, the functionload_image_batch
applies a transformation at line 201 ,img_tensor = (img_tensor / 255. - 0.5) * 2
. I wonder whether or not there is a need to apply this at the inference code. Many thanks for you time!