deep-high-resolution-net.pytorch
deep-high-resolution-net.pytorch copied to clipboard
'center' and 'scale' in MPII annotation
Hi! How can I get the 'center' and 'scale' in MPII annotation? Are these values obtained by a detector? Or they're just the average of the joints coordinates?
@violetnotfat i think the author has computed it in json file, but, i am not sure whether the json file contains single-person or multi-person, further, they did not provide how they computed it
I calculated the averages of the all subsets of joint coordinates and compared between the values and "center" values. However, they aren't always equal to the "center" values. I think "center" values are annotated manually.
I calculated the averages of the all subsets of joint coordinates and compared between the values and "center" values. However, they aren't always equal to the "center" values. I think "center" values are annotated manually.
manually? in addition, the number of smaple between official test.h5 and test.json of HRNet is different . so i use offical test.h5 file to test, and can the difference of center and scale influence the submitted result ?