Authors: Andrei Niculae, Andy Catruna, Adrian Cosma, Daniel Rosner, Emilian Radoi

Published on: April 18, 2024

Impact Score: 7.2

Arxiv code: Arxiv:2404.12183

Summary

• What is new: Introduction of a task-targeted artifact correction model designed to pre-process and enhance surveillance footage for gait analysis without degrading performance on high-quality data.
• Why this is important: Low quality and high noise levels in surveillance footage significantly impact the accuracy of pose estimation algorithms, crucial for gait analysis.
• What the research proposes: A processing pipeline that includes a newly developed artifact correction model to improve surveillance footage before pose estimation, optimized to work with HRNet.
• Results: Improved pose estimation accuracy on low-quality surveillance footage while maintaining performance on high-quality footage, leading to enhanced gait analysis results.

Technical Details

Technological frameworks used: nan

Models used: HRNet, Task-targeted artifact correction model

Data used: Automatically annotated low-quality videos for training artifact correction model

Potential Impact

Security and surveillance companies, healthcare providers using surveillance for patient monitoring, and companies developing gait analysis technology could benefit from these insights.

Want to implement this idea in a business?

We have generated a startup concept here: StrideScan.