Research papers · Posted 2025.10.17 · Publication / acceptance: author-reported
ObjectTransforms for Uncertainty Quantification and Reduction in Vision-Based Perception for Autonomous Vehicles
ObjectTransforms for Uncertainty Quantification and Reduction in Vision-Based Perception for Autonomous Vehicles
Overview
ObjectTransforms uses object-specific augmentation during training and perturbation-based score variation during inference. Experiments evaluate uncertainty reduction, false-positive filtering and missed-detection recovery in vision-based automotive object detection.
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.
Accepted at International Conference on Computer Vision (ICCV) 2025 Workshops
Version: arXiv v1 — manuscript revision