STVIR / pysot

SenseTime Research platform for single object tracking, implementing algorithms like SiamRPN and SiamMask.
Apache License 2.0
4.43k stars 1.1k forks source link

mask_loss is equal to None? #282

Open x1aoo opened 4 years ago

x1aoo commented 4 years ago

Hi, In /pysot/pysot/models/model_builder.py line 112, the mask_loss is always equal to None and the output of "total loss" will be int * NoneType value, it will have a error:

outputs['total_loss'] += cfg.TRAIN.MASK_WEIGHT * mask_loss
TypeError: unsupported operand type(s) for *: 'int' and 'NoneType'
ZhiyuanChen commented 4 years ago

Sorry, we haven't finished this part yet. Pls set the cfg.MASK.MASK to False

dudiaoxifenglie commented 3 years ago

Set the cfg.MASK.MASK to False? Is it still a SIAMASK algorithm?

Radishzinski commented 2 years ago

将 cfg.MASK.MASK 设置为 False?还是SIAMASK算法?

Hi, I'm training the siammask model with my own data and I'm having the same problem during the run. Can you please tell me how you solved it.