oujieww

@SteveCruz thanks for your imformation. I change my code to : model.net.Sub([blob_out, 'label'],'diff') model.net.Abs(['diff'],'adiff') model.net.DotProduct(['adiff','adiff'],'dist') model.net.AveragedLoss(['dist'],'final_loss') and it work.but i get very big loss . if loss still so big...

@SyGoing no , my finall loss is too big ,i still try to change lr .but ~~~~~,i have no idea ~~~~, how do u think about this?

@SyGoing i think should have multi stage , i add normlize for net, my last problem is solved. but ~~~~finall loss is still too big~~~,and result is not good

@SyGoing sorry not normlize, should be initialize

hi，do u solve it？I also want to know it.pls！

I also have this problem,I want to know can i just make my dataset with 'input_image/3d_rgb' and '16_heatmaps as label' to setup my dataset.maybe this way will be different with...

@feitiandemiaomi oho.yes,did u use lmbd?there too many other labels needed in author's model.I want to change some thing ,maybe to setup my dataset just with 'data_image' and 'label_16_heatmaps'.I want to...

i thank label is same as heatmaps，u can see them when u run demo.

@wwxzxd the digits in LimbSeq are order of heatmap mapping to jionts. for example ,[2,3] maybe mean the limb of left upper arm,so [31,32] should mean the PAF of left...

and how about the time for "cuda heatmaps for instance transformed to keypoints coordinate"