Segment-Level Road Obstacle Detection Using Visual Foundation Model Priors and Likelihood Ratios

2024-12-07

arXiv.org

AI Summary

To address poor robustness, threshold sensitivity, and fragmented predictions in obstacle detection for dynamic road scenes in autonomous driving, this paper proposes the first end-to-end segment-level obstacle detection framework. Departing from conventional pixel-wise scoring and threshold-based segmentation, our method innovatively integrates semantic segment features from vision foundation models (e.g., ViT, CLIP) with a Bayesian likelihood ratio test (LRT) mechanism, enabling threshold-free, segment-level obstacle discrimination. We further enhance generalization and robustness through superpixel-guided semantic segment aggregation and prior-driven LRT modeling. Evaluated on RoadObstacle and LostAndFound benchmarks, our approach achieves state-of-the-art performance, significantly reducing false positive rates while yielding more coherent and reliable detections—eliminating reliance on hand-crafted thresholds entirely.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Detecting road obstacles is essential for autonomous vehicles to navigate dynamic and complex traffic environments safely. Current road obstacle detection methods typically assign a score to each pixel and apply a threshold to generate final predictions. However, selecting an appropriate threshold is challenging, and the per-pixel classification approach often leads to fragmented predictions with numerous false positives. In this work, we propose a novel method that leverages segment-level features from visual foundation models and likelihood ratios to predict road obstacles directly. By focusing on segments rather than individual pixels, our approach enhances detection accuracy, reduces false positives, and offers increased robustness to scene variability. We benchmark our approach against existing methods on the RoadObstacle and LostAndFound datasets, achieving state-of-the-art performance without needing a predefined threshold.

Problem

• Autonomous Vehicles
• Obstacle Detection
• Robustness

Innovation

• Obstacle Detection
• Large Patch Information
• Mathematical Techniques