Reproduce “SAFe: Segment-guided Aggregation of Feature Densities for Anomaly-aware Segmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Anja Delić, Jurica Runtas, Marin Oršić, Ivan Marković, Ivan Petrović
SAFe detects unfamiliar regions in road scenes by estimating class-conditional feature likelihoods with frozen DINOv3 ViT and ConvNeXt backbones, fusing their evidence, and refining scores with SAM3 masks. On SMIYC ObstacleTrack, it reaches 95.5 AP and 0.04 FPR95, versus RbA’s 95.1 and 0.1. This result is limited to that benchmark and task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.