siyuhuang / tod Goto Github PK

Thanks a lot for sharing codes of your interesting work. I wonder how to calculate and generate the Figure. 1? Would you like to share the codes about it?
Besides, I have two another questions:
(1) According to Corollary 2, it needs to consider the factors of eta and C (const) to estimate the losses, but it seems no related parameter in the codes. Does it show any influence of performance about it?
(2) Did you tested and employed TOD on general semi-supervised learning task?
Many thanks.

Hello, thank you for this nice work.

I am trying to implement the active learning on Cityscapes with DRN model (https://github.com/fyu/drn). I followed the implementation details in your paper: drn_d_22 architecture, Adam optimizer, lr 5e-4, epoch 40, batch size 4, crop size 688. However, the performance already reaches 59.26 mIoU over 30% samples selected randomly, which is much higher than your reported performance in Fig. 8. Could you kindly tell me whether I made anything wrong? or can you share your command to train the drn network?

Thank you very much.

虽然我知道此时model['cod']是随机初始化的，model['backbone']是经过第一轮训练的

Hi, Siyu Huang:

Thanks for your nice work. I learned a lot from your theoretical analysis in part 3. After carefully reading your paper, I found there is a miss of ECCV 2020 paper which is similar to your work. Both borrow ideas from semi-supervised learning. What's more, I am curious about the comparison and analysis between your work and the above ECCV paper under the same setting.

Best wishes.

I am very interested in reading your research. I would like to ask if you can open source the code for the Cityscapes dataset experiment.

Hello! I want to know how to update the parameters of models['cod'].

I'm looking forward to your reply！

Hi, thanks for your work. However, this code seems to be only available on classification datasets. How can I run TOD on Cityscapes?