Semi-Supervised Pipe Video Temporal Defect Interval Localization

TL;DR

This paper presents PipeSPO, a semi-supervised method for accurately localizing defects in sewer pipe videos using minimal annotations, leveraging visual odometry and unlabeled data to outperform existing approaches.

Contribution

The paper introduces PipeSPO, a novel semi-supervised approach that combines visual odometry and weak supervision for defect localization in pipe videos, reducing annotation costs.

Findings

Achieves 41.89% average precision across IoU thresholds 0.1-0.7.

Improves performance by 8.14% over state-of-the-art methods.

Effectively leverages unlabeled data and weak supervision for defect localization.

Abstract

In sewer pipe Closed-Circuit Television (CCTV) inspection, accurate temporal defect localization is essential for effective defect classification, detection, segmentation and quantification. Industry standards typically do not require time-interval annotations, even though they are more informative than time-point annotations for defect localization, resulting in additional annotation costs when fully supervised methods are used. Additionally, differences in scene types and camera motion patterns between pipe inspections and Temporal Action Localization (TAL) hinder the effective transfer of point-supervised TAL methods. Therefore, this study introduces a Semi-supervised multi-Prototype-based method incorporating visual Odometry for enhanced attention guidance (PipeSPO). PipeSPO fully leverages unlabeled data through unsupervised pretext tasks and utilizes time-point annotated data with a weakly supervised multi-prototype-based method, relying on visual odometry features to capture camera pose information. Experiments on real-world datasets demonstrate that PipeSPO achieves 41.89% average precision across Intersection over Union (IoU) thresholds of 0.1-0.7, improving by 8.14% over current state-of-the-art methods.