search

seoungwugoh / RGMP

Fast Video Object Segmentation by Reference-Guided Mask Propagation
192 stars 37 forks source link

the problem of train.py #18

Open XiandaGuo opened 4 years ago

XiandaGuo commented 4 years ago

why do not you use bptt_hsm in train.py? what is the meaning of ntokens in line 134 of train.py? looking forward your replay!Thank you very much

Starboy-at-earth commented 4 years ago

I have encountered the same problem. Look forward to that the author can give some explanation, thank you in advance.

zgf6781882 commented 4 years ago

I've run the training code, but I get very bad results, so I always encounter the following problems. Traceback (most recent call last): File "train.py", line 301, in writer.add_scalar('Train/IOU', iou(torch.cat((1-all_E, all_E), dim=1), all_M), i + epoch * iters_per_epoch) File "/_guifangzhang/vos/RGMP/utils.py", line 93, in iou return i / u ZeroDivisionError: float division by zero

Does anyone encounter the same situation???

gbc-sid commented 3 years ago

I have the same iou problem and the segment results are bad. Have you solved this problem?