Supporting Safety Analysis of Image-processing DNNs through Clustering-based Approaches

Mohammed Oualid Attaoui, Hazem Fahmy, Fabrizio Pastore, Lionel Briand

Research output: Contribution to journalArticlepeer-review

Abstract

The adoption of deep neural networks (DNNs) in safety-critical contexts is often prevented by the lack of effective means to explain their results, especially when they are erroneous. In our previous work, we proposed a white-box approach (HUDD) and a black-box approach (SAFE) to automatically characterize DNN failures. They both identify clusters of similar images from a potentially large set of images leading to DNN failures. However, the analysis pipelines for HUDD and SAFE were instantiated in specific ways according to common practices, deferring the analysis of other pipelines to future work. In this article, we report on an empirical evaluation of 99 different pipelines for root cause analysis of DNN failures. They combine transfer learning, autoencoders, heatmaps of neuron relevance, dimensionality reduction techniques, and different clustering algorithms. Our results show that the best pipeline combines transfer learning, DBSCAN, and UMAP. It leads to clusters almost exclusively capturing images of the same failure scenario, thus facilitating root cause analysis. Further, it generates distinct clusters for each root cause of failure, thus enabling engineers to detect all the unsafe scenarios. Interestingly, these results hold even for failure scenarios that are only observed in a small percentage of the failing images.

Original languageEnglish
Article number130
JournalACM Transactions on Software Engineering and Methodology
Volume33
Issue number5
DOIs
Publication statusPublished - 3 Jun 2024

Keywords

  • clustering
  • DNN debugging
  • DNN explanation
  • DNN functional safety analysis
  • transfer learning

Fingerprint

Dive into the research topics of 'Supporting Safety Analysis of Image-processing DNNs through Clustering-based Approaches'. Together they form a unique fingerprint.

Cite this