About setting of baseline

[zhizhangxian]

Hi，your work is so amazing and help me a lot!
I mentioned in your reported result in paper, you did some comparison between deeplab and your results on different datasets, but I ran your code with batch size=16, deeplab v3+ model , resnet101 backbone on VOC2012 dataset, but only get 0.772miou after about 30k iterations(just the default setting), so can you tell me how can I set the hyper-parameters to get the desire result (including 78.8 with crossentropy loss and higer miou with your proposed loss)?
this is my result on val dataset:

Thank you very much for your excellent work!

[Shuai Zhao]

Glad to hear that our work can benefit someone.

In fact, 78.8% is the data report in the official paper, i.e., the DeepLabv3+ paper.
Our reimplementation result is 78.17% with DeepLabv3+ (batch_size=16, output_stride=16).
During my experiments, I also had some similar results as you with DeepLabv3+.
So I do not have many tricks to help you.

But you may find something helpful in this paper https://arxiv.org/abs/1812.01187.
It helped me a lot when I took my experiments.

[zhizhangxian]

Thanks very much！