CONTA

CONTA copied to clipboard

pseudo_mask

8

Thanks for your work. How to replace the ground_truth annotation in PASCAL VOC 2012 with the generated pseudo_mask?

labiao

About the Context M and C SET

2

The set generation of average masks and the composition of M are in persudo-mask, sege is a separate semantic segmentation model

Wzhsgsg

About the Context M

1

Thanks for the interesting work. The paper says that \[X; M_t\] are concatenated to train the classifier. Could you point out where they are concatenated in the code? I am...

VelsLiu

Hi~ Thanks for your great codes! The model SEAM+CONTA, uses deeplabv2 with resnet101 as the final segmentation network, but the SEAM uses deeplabv1 with resnet38, is this comparison fair?

pigcv

TyroneLi

When attempting to execute main.py a pickling error is returned and after tracing it back it is `state_dict = torch.load(CONFIG.MODEL.INIT_MODEL)` line that is throwing the error. Also, the INIT_MODEL in...

gmvincent

Question about Equation 3 in the main paper

2

Hi, I wonder whether you could explain why W1 and W2 are learnable parameters since there is no objective function specified in step 4. Thanks

IN-Gamut

Question about training process and "rounds"

1

I'm confused about your paper's use of the term "round" and your training scheme in general. Is "round" the same thing as your term t? Am I correct in understanding...

lambertwx

How to use the "m_t.py"

1

Thanks for your excellent work. I met some troubles when I try to write scripts to use the "m_t.py", which is the implementation of the Eq.(3) in your article. (1)...

cz-xjtu

TyroneLi