A reusable framework generates verification instances with provably known robustness labels, revealing numeric tolerance issues and bugs in five verifiers while introducing difficulty profiles to diagnose failure modes.
A systematic review of robustness in deep learning for computer vision: Mind the gap?
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
representative citing papers
SegSTRONG-C provides a new benchmark where top models reach 0.9394 DSC and 0.9301 NSD on corrupted surgical tool segmentation tests, showing conventional techniques help but calling for more innovative robustness methods.
An interdisciplinary workshop produced a catalog of ideas and a roadmap showing how meta-research can tackle challenges like reproducibility, transparency, and ethical implementation in trustworthy AI for healthcare.
citing papers explorer
-
Stress-Testing Neural Network Verifiers with Provably Robust Instances
A reusable framework generates verification instances with provably known robustness labels, revealing numeric tolerance issues and bugs in five verifiers while introducing difficulty profiles to diagnose failure modes.
-
SegSTRONG-C: Segmenting Surgical Tools Robustly On Non-adversarial Generated Corruptions -- An EndoVis'24 Challenge
SegSTRONG-C provides a new benchmark where top models reach 0.9394 DSC and 0.9301 NSD on corrupted surgical tool segmentation tests, showing conventional techniques help but calling for more innovative robustness methods.
-
Advancing Trustworthy AI in Healthcare Through Meta-Research: Results of an Interdisciplinary Design-Thinking Workshop
An interdisciplinary workshop produced a catalog of ideas and a roadmap showing how meta-research can tackle challenges like reproducibility, transparency, and ethical implementation in trustworthy AI for healthcare.