Yuliang (Zack) Zou

You can check this [thread](https://github.com/googleinterns/wss/issues/5)

You can check this [file](https://github.com/googleinterns/wss/blob/master/third_party/deeplab/datasets/data_generator.py)

Yeah sure. Thank you so much~

Hi Carl, any updates on the models?

@rvarun7777 Can you share your pretrained weights?

Hi, thanks for the PR. Can you briefly describe why the current performance evaluation is inaccurate?

Some of the code is related to a specific version of cuda. As far as I know, TF1.8 is not using that version. If you really want FlowNet2, I would...

Which TF version are you using? And maybe you can check [this](https://github.com/kaldi-asr/kaldi/issues/1898)

@Timthony You need to use 1.2. We don't support other versions currently.

Did you the empty GPU or other occupied GPUs? Or did you load your custom high-resolution image?