SAFE: Safety Analysis and Retraining of DNNs

Mohammed Oualid Attaoui, Fabrizio Pastore, Lionel Briand

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We present SAFE, a tool based on a black-box approach to automatically characterize the root causes of Deep Neural Network (DNN) failures. SAFE relies on VGGNet-16, a transfer learning model pre-trained on ImageNet, to extract the features from errorinducing images. After feature extraction, SAFE applies a densitybased clustering algorithm to discover arbitrarily shaped clusters of images modeling plausible causes of failures. By relying on the identified clusters, SAFE can select a set of additional images to be used to retrain and improve the DNN efficiently. Empirical results show the potential of SAFE in identifying different root causes of DNN failures based on case studies in the automotive domain. It also yields significant improvements in DNN accuracy after retraining while saving considerable execution time and memory compared to alternatives. A demo video of SAFE is available at https://youtu.be/8QD-PPFTZxs.

Original languageEnglish
Title of host publicationProceedings - 2024 ACM/IEEE 46th International Conference on Software Engineering
Subtitle of host publicationCompanion, ICSE-Companion 2024
PublisherIEEE Computer Society
Pages74-78
Number of pages5
ISBN (Electronic)9798400705021
DOIs
Publication statusPublished - 14 Apr 2024
Externally publishedYes
Event46th International Conference on Software Engineering: Companion, ICSE-Companion 2024 - Lisbon, Portugal
Duration: 14 Apr 202420 Apr 2024

Publication series

NameProceedings - International Conference on Software Engineering
ISSN (Print)0270-5257

Conference

Conference46th International Conference on Software Engineering: Companion, ICSE-Companion 2024
Country/TerritoryPortugal
CityLisbon
Period14/04/2420/04/24

Keywords

  • DNN Debugging
  • DNN explanation
  • Functional Safety Analysis

Fingerprint

Dive into the research topics of 'SAFE: Safety Analysis and Retraining of DNNs'. Together they form a unique fingerprint.

Cite this