Scalable Semi-Automatic Annotation for Multi-Camera Person Tracking.

Area of Science:

• Computer Vision
• Machine Learning
• Video Analysis

Background:

• Evaluating multi-camera people-tracking systems requires accurate position annotations on large video datasets.
• Manual annotation is labor-intensive and time-consuming, hindering scalability.
• Existing methods lack robustness and efficiency for complex tracking scenarios.

Purpose of the Study:

• To propose a generic, semi-automatic methodology for generating reliable position annotations for multi-camera people-tracking evaluation.
• To reduce the manual effort required for annotating large video datasets.
• To provide a framework adaptable to new datasets and tracking algorithms.

Main Methods:

• A semi-automatic annotation framework combining automated consensus tracking and human verification.
• Automated annotation by estimating consensus tracking results from multiple trackers and detectors.
• Human verification of unreliable tracks through a fast binary decision task.

Main Results:

• Achieved 80% automatic annotation for frames with 60cm tracking accuracy on a 6-hour dataset.
• Required only 2.4 hours of manual labor for the remaining 20% of frames.
• Demonstrated 99% accuracy for automatically annotated frames after visual inspection.

Conclusions:

• The proposed methodology offers an efficient and reliable solution for generating position annotations for people-tracking evaluation.
• The framework is generic, adaptable to new datasets, and can incorporate additional trackers.
• The study provides guidelines and an exploratory analysis for multi-target tracking scenarios.