sairin1202 / semantic-aware-attention-based-deep-object-co-segmentation Goto Github PK

View Code? Open in Web Editor NEW

68.0 68.0 17.0 2.13 MB

Semantic Aware Attention Based Deep Object Co-segmentation

Python 100.00%

attention co-segmentation deep-learning pytorch segmentation

semantic-aware-attention-based-deep-object-co-segmentation's People

Contributors

Stargazers

Watchers

Forkers

amieefu benjaminzhouyj silliam wandoucao mismabs avinash-chouhan jingweiqu wangsff kartheekmedathati hopstone prretty varshnidimpy choodly pixiedust18 djx2726889 dengjl-hub gnkgo

semantic-aware-attention-based-deep-object-co-segmentation's Issues

What is the setting of your group co-segmentation experiments? How to construct the group co-segmentation dataset? randomly or pair all the images ? In your paper, I could not find the size of your group .Did you do the experiments with random group size or fixed group size? Thank you.

Is the best performance got at epoch1 iter 5000? Thanks!

I am also interested in the pre-trained model.

Hi, @sairin1202

I am also interested in the pre-trained model.
Would you please share it to me?

My mail:
[email protected]

Thanks in advance~
Best,
@bemoregt.

Can you share your ACCV paper?

Hi, I am very interested in your ACCV work, however I can not find the paper. So could you provide a camera ready pdf file? [email protected], thanks

ReadMe Drive Link Doesn't Work

Hiya, the drive link for preprocessing dataset for the ReadMe does not work. (Unless you intend for everyone who wants to download to the dataset to request access).

Also would you be willing to release the pretrained weights for your network?

Training is very slow

The training is very slow.
Even the evaluate part takes hours to finish.

How much time did it take to train the model?

Model saving error happened.

HI, @sairin1202

I have met this error when run train.py.

"---------------------------------------------
epoch0 iter 0/5300 BCE loss:
testing......
pixel accuracy: 0.18804569360921725 correct: 4631072796 wrong: 19996307428
precision: 0.18803271531914503 true_positive: 4630611028 positive: 24626624256
jaccard score: 0.18803046902816284 intersection: 4630611028 union: 24626918456

epoch0 iter 2000/5300 BCE loss:
testing......
pixel accuracy: 0.9180765624825236 correct: 22609820579 wrong: 2017559645
precision: 0.8025394126087568 true_positive: 3466181168 positive: 4319016753
jaccard score: 0.6320833325643175 intersection: 3466181168 union: 5483740813
Traceback (most recent call last):
File "train.py", line 183, in
trainer.train()
File "train.py", line 178, in train
'epoch{}iter{}.pkl' % (epoch, i))
TypeError: not all arguments converted during string formatting"

What's wrong to me ?

Thanks in advance ~.
Best,
@bemoregt.

The trained model

Can you provide a trained model? Thank you

Question on a paper

I cannot find out a paper concerning your code.
Can I ask you the link of the paper?

thank you.

Pretrained model

你好,
我正在做另一个与你类似的项目。
我想问一下，你能不能把预先训练好的模型上传呢？
非常感谢你的帮忙.
Email: [email protected]

Pretrained Model

Hi. Thanks for providing the code. Can you also upload the pre-trained model, it will be helpful. I currently do not have enough resources (I mostly use Google Colab) to train the entire network from scratch, so it will be grateful if you upload the pre-trained network.