search

caizhongang / SMPLer-X

Official Code for "SMPLer-X: Scaling Up Expressive Human Pose and Shape Estimation"
https://caizhongang.github.io/projects/SMPLer-X/
Other
1.01k stars 73 forks source link

random output on inference with pretrained model #8

Closed noeykan closed 1 year ago

noeykan commented 1 year ago

Hi,

I've tested inference with your pretrained model, but the result seems random value. As I tried to debug this, input image cropped was okay.

000471

I've tested all pretrained models but failed.

Could you check them again?

caizhongang commented 1 year ago

Hi @noeykan , this doesn't seem right. Have you checked if you loaded the pretrained model successfully? PyTorch may not issue a warning even if the loading fails (e.g., file not found). Please also post your inference log.

noeykan commented 1 year ago

@caizhongang Oh, I fixed this problem. It was my mistake. Because of some gpu errors from my computer, I changed the code in common/base.py as below at first and it made the random results as I posted before.

class Demoer(Base): ...

prepare network

     self.logger.info("Creating graph...")
     model = get_model('test')
     model = DataParallel(model).cuda()
     # model = DataParallel(model).cuda()
     model = model.cuda()
     ckpt = torch.load(cfg.pretrained_model_path)
     ...

I thought these changes had not affected the result, but when I rolled back my modifications, I could see the right results outputted from the inference pipeline.

Thank you.