Epistemic and aleatoric uncertainty quantification for crack detection using a Bayesian Boundary Aware Convolutional Network

Published in Reliability Engineering & System Safety, 2023 · arXiv

Representative figure for this publication
Crack detections on the CrackForest and DeepCrack datasets: input image, prediction, ground truth, and the epistemic and aleatoric uncertainty maps produced by the Bayesian boundary aware network.

Separates epistemic from aleatoric uncertainty in neural-network crack detection, enabling risk-aware structural inspection.

Representative publication of this research line.

Recommended citation

@article{rathnakumar2023epistemic,
  title = {Epistemic and aleatoric uncertainty quantification for crack detection using a Bayesian Boundary Aware Convolutional Network},
  author = {Rathnakumar, R. and Pang, Y. and Liu, Y.},
  journal = {Reliability Engineering \& System Safety},
  pages = {109547},
  year = {2023},
  eprint = {2302.06827},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
  url = {https://arxiv.org/abs/2302.06827}
}