windvchen / inr-harmonization Goto Github PK

I didn't find anything about the speed of reasoning.What is the speed of inference

I was working on to convert onnx but getting too many error , @WindVChen
If possible you can define reference input and output model params because i have seen the model taking list for all image,mask,fg_coordinates but this are depent on image size so possible if will change the split_num then we will get something else input size for fg_coordinates, but you mention for loop for to get all var from datasets class that is taking the all image in dict formats having images in array .
My issue is that while taking all of static list all composite ,mask,fg_coordniate,all are in list of size 4 and having tensor with [B,C,W,H].
if possible can you help me out

efficient_inference_for_square_image.py is

Hi,
It helps to save memory by spliting image into patches, but each patches are treated individually, is there another way to ultilize the global info of the original image for harmonization with limited GPUs?
Another question is, I'm trying to do video harmonization, sometimes the results of adjacent frames are different while the background is quite similar. Also, the same thing happens in many other works. Is it an unsolved problem for image harmonization?

cannot download HRNetV2 weight