ROM and RUM

Unofficial implementation of "Rethinking Semantic Segmentation Evaluation for Explainability and Model Selection".

@article{Zhang2021RethinkingSS,
  title={Rethinking Semantic Segmentation Evaluation for Explainability and Model Selection},
  author={Yuxiang Zhang and Sachin Mehta and Anat Caspi},
  journal={ArXiv},
  year={2021},
  volume={abs/2101.08418}
}

This repository contains the following:

• Region-wise over-segmentation measure (ROM)
• Region-wise under-segmentation measure (RUM)

TODO:

• Multi-class function (currently binary evaluation function; I just iterate through all the classes)
• Add test code
• Add better tests
• More Efficient Implementation