What is the inference speed for e.g. semantic segmentation using 1024x1024 (referring to table 5)? Measured on GPU of your choice, just to get a feeling?
HRNet
