Introduction

This is a project which just move the EM-Adapt to tensorflow. The EM-Adapt is referring to the approach for weakly-supervised semantic segmentation in the paper "Weakly- and semi- supervised learning of a DCNN for semantic image segmentation". And here, I just use the tensorflow to implement the approach with the help of the published code.

Citing this repository

If you find this code useful in your research, please consider citing them:

@inproceedings{papandreou15weak,

​ title={Weakly- and Semi- Supervised Leaning of a DCNN for Semantic Image Segmentation},

​ author={George, Papandreou and Liang-Chieh Chen and Kevin Murphy and Alan L Yuille},

​ journal={arxiv:1502,02734},

​ year={2015}

}

Preparation

for using this code, you have to do something else:

1. Download the data and model

1. for pascal data, please referring to its official website and to the augmental SBD data. Just download it and extract it in the ./, then run convert.py using 'python convert.py'.
2. for the init.model, please referring to EM-ADAPT or google driver. And download it and extract in the model/ .

For more details, you can referring to the correspond code files or leave a message in the issue.

Training

then, you just input the following sentence to train it.

python deeplab.py <gpu_id>

Result

the final result on the validation dataset of pascal voc 2012 is 37.98% miou while it is 38.2% in the paper. Note that we use the crf while test the trained model, and you can look through my other project to see how to perform densecrf using python.

Evaluation

I just release a project to provide the code for evaluation.