Road Obstacle Video Segmentation

2025-09-16

AI Summary

Existing road obstacle segmentation methods predominantly operate on single-frame images, neglecting temporal consistency and thereby suffering from severe prediction flickering in long videos. This work first systematically establishes the video-level temporal nature of this task and constructs four novel long-sequence video benchmarks to comprehensively evaluate the temporal robustness of both image- and video-based segmentation methods. Leveraging vision foundation models, we propose two strong baselines that explicitly model inter-frame consistency via spatiotemporal feature fusion. Our approaches achieve state-of-the-art performance across multiple video segmentation benchmarks, significantly improving both segmentation accuracy and temporal stability. We release the first dedicated video segmentation benchmark for road obstacles—ROAD-Video—thereby establishing a new paradigm for temporal modeling in this domain. This work provides a standardized evaluation framework and a principled technical pathway for video semantic segmentation in autonomous driving scenarios.

Technology Category

• Computer Vision: Segmentation
• Planning, Routing, and Scheduling: Temporal Planning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

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
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

With the growing deployment of autonomous driving agents, the detection and segmentation of road obstacles have become critical to ensure safe autonomous navigation. However, existing road-obstacle segmentation methods are applied on individual frames, overlooking the temporal nature of the problem, leading to inconsistent prediction maps between consecutive frames. In this work, we demonstrate that the road-obstacle segmentation task is inherently temporal, since the segmentation maps for consecutive frames are strongly correlated. To address this, we curate and adapt four evaluation benchmarks for road-obstacle video segmentation and evaluate 11 state-of-the-art image- and video-based segmentation methods on these benchmarks. Moreover, we introduce two strong baseline methods based on vision foundation models. Our approach establishes a new state-of-the-art in road-obstacle video segmentation for long-range video sequences, providing valuable insights and direction for future research.

Problem

• Addressing inconsistent road obstacle segmentation across video frames
• Evaluating temporal methods for road obstacle video segmentation
• Establishing benchmarks for video-based obstacle detection consistency

Innovation

• Video-based segmentation for temporal consistency
• Vision foundation models for baseline methods
• Long-range video sequences for obstacle segmentation