Label Correction for Road Segmentation Using Road-side Cameras

2025-02-03

AI Summary

To address the high annotation cost for all-weather road segmentation—particularly under winter conditions—this paper proposes a lightweight, semi-automatic annotation paradigm tailored for roadside cameras: only one frame per camera requires manual annotation; high-precision label transfer is achieved via frequency-domain image registration to compensate for minor viewpoint shifts. This approach drastically reduces annotation effort and eliminates the need for large-scale multi-weather video collection and labeling. Evaluated on real-world video data from 927 roadside cameras in Finland, the method integrates semantic segmentation models (e.g., U-Net, DeepLabv3+) and undergoes rigorous cross-domain generalization assessment. Results show an 8.2% improvement in mean Intersection-over-Union (mIoU) within the roadside domain and strong robustness in cross-domain evaluation using vehicle-mounted cameras, validating both the quality of generated labels and the generalization capability of downstream models.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Transportation

Application Category

• Semantics and Knowledge: Representation, semantic annotation, enhancement, enrichments, access and/or integration of a variety of data on the Web
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Reliable road segmentation in all weather conditions is critical for intelligent transportation applications, autonomous vehicles and advanced driver's assistance systems. For robust performance, all weather conditions should be included in the training data of deep learning-based perception models. However, collecting and annotating such a dataset requires extensive resources. In this paper, existing roadside camera infrastructure is utilized for collecting road data in varying weather conditions automatically. Additionally, a novel semi-automatic annotation method for roadside cameras is proposed. For each camera, only one frame is labeled manually and then the label is transferred to other frames of that camera feed. The small camera movements between frames are compensated using frequency domain image registration. The proposed method is validated with roadside camera data collected from 927 cameras across Finland over 4 month time period during winter. Training on the semi-automatically labeled data boosted the segmentation performance of several deep learning segmentation models. Testing was carried out on two different datasets to evaluate the robustness of the resulting models. These datasets were an in-domain roadside camera dataset and out-of-domain dataset captured with a vehicle on-board camera.

Problem

• Road Recognition
• Weather Variability
• Data Annotation

Innovation

• Automated Image Annotation
• Robust Road Recognition
• Weather-Adaptive Data Collection