Augmentation Consistency-Guided Self-Training for Source-Free Domain Adaptive Semantic Segmentation

Abstract

We focus on source-free domain adaptation for semantic segmentation, wherein a source model must adapt itself to a new target domain given only unlabeled target data. We propose Augmentation Consistency-guided Self-training (AUGCO), an adaptation algorithm that uses the model's pixel-level predictive consistency across diverse, automatically generated views of each target image along with model confidence to identify reliable pixel predictions, and selectively self-trains on those, leading to state-of-the-art performance within a simple to implement and fast to converge approach.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Prabhu et al. "Augmentation Consistency-Guided Self-Training for Source-Free Domain Adaptive Semantic Segmentation." NeurIPS 2022 Workshops: DistShift, 2022.

Markdown

[Prabhu et al. "Augmentation Consistency-Guided Self-Training for Source-Free Domain Adaptive Semantic Segmentation." NeurIPS 2022 Workshops: DistShift, 2022.](https://mlanthology.org/neuripsw/2022/prabhu2022neuripsw-augmentation/)

BibTeX

@inproceedings{prabhu2022neuripsw-augmentation,
  title     = {{Augmentation Consistency-Guided Self-Training for Source-Free Domain Adaptive Semantic Segmentation}},
  author    = {Prabhu, Viraj Uday and Khare, Shivam and Kartik, Deeksha and Hoffman, Judy},
  booktitle = {NeurIPS 2022 Workshops: DistShift},
  year      = {2022},
  url       = {https://mlanthology.org/neuripsw/2022/prabhu2022neuripsw-augmentation/}
}