Research papers · Posted 2025.11.08 · Publication / acceptance: author-reported
Runtime Safety Monitoring of Deep Neural Networks for Perception: A Survey
Runtime Safety Monitoring of Deep Neural Networks for Perception: A Survey
Overview
This survey organizes perception-DNN runtime monitors by inputs, internal representations and outputs. It maps methods to generalization errors, distribution shifts and adversarial attacks, comparing strengths, limitations and open research challenges.
Publication and version information
Publication, acceptance or venue information is author-reported on arXiv. The exact source wording is preserved below.
Author-supplied source information
Author-supplied information on arXiv.
6 pages, 1 figure, 2 tables, accepted at IEEE SMC 2025 in Vienna, presented on 8th October 2025
Version: arXiv v1 — manuscript revision