LOSS FUNCTIONS IN THE ERA OF SEMANTIC SEGMENTATION: A SURVEY AND OUTLOOK
Reza Azad, Moein Heidari, Kadir Yilmaz, Michael Hüttemann, Sanaz Karimijafarbigloo, Yuli Wu, Anke Schmeink, Dorit Merhof

Abstract: Semantic image segmentation, the process of classifying each pixel in an image into a particular class, plays an important role in many visual understanding systems. In the context of deep learning-based segmentation algorithms, the choice of an appropriate loss function is crucial for training the model effectively. It quantifies the difference between the predicted segmentation and the ground truth, providing a measure of how well the model is performing. To aid researchers in identifying the optimal loss function for their particular application, this survey provides a comprehensive and unified review of $25$ loss functions utilized in image segmentation. We provide a novel taxonomy and thorough review of how these loss functions are customized and leveraged in image segmentation, with a systematic categorization emphasizing their significant features and applications. Furthermore, to evaluate the efficacy of these methods in real-world scenarios, we propose unbiased evaluations of renowned loss functions on established medical and natural image datasets. We conclude this review by identifying current challenges and unveiling future research opportunities.

🔥🔥 Our survey paper on arXiv: Loss Functions in the Era of Semantic Segmentation: A Survey and Outlook🔥🔥

A curated list of awesome loss functions in semantic segmentation. This repo supplements our survey paper. We intend to continuously update it. We strongly encourage authors of relevant works to make a pull request and add their paper's information.

If you find our work useful in your research, please consider citing:

@misc{azad2023loss,
      title={Loss Functions in the Era of Semantic Segmentation: A Survey and Outlook}, 
      author={Reza Azad and Moein Heidary and Kadir Yilmaz and Michael Hüttemann and Sanaz Karimijafarbigloo and Yuli Wu and Anke Schmeink and Dorit Merhof},
      year={2023},
      eprint={2312.05391},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}