Reproduce “FROG: A new people detection dataset for knee-high 2D range finders” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fernando Amodeo, Noé Pérez-Higueras, Luis Merino, Fernando Caballero
Researchers from the Service Robotics Laboratory at Universidad Pablo de Olavide introduced FROG, a new comprehensively annotated dataset for people detection using knee-high 2D LiDARs, which is 17 times larger than previous resources. They also proposed LFE-PPN, an end-to-end deep learning detector achieving real-time inference at over 500 Hz, significantly faster than existing methods while maintaining competitive accuracy.
Key insights
- A comprehensive and high-quality 2D LiDAR dataset, like FROG, is essential for training robust deep learning models and facilitating standardized research in people detection for mobile robotics.
- End-to-end deep learning architectures can effectively process raw 2D laser scan data, offering substantial inference speed advantages (over 500 Hz) suitable for real-time robotic applications, even if slightly below the highest performing temporal methods in accuracy.
- While temporal aggregation (e.g., DR-SPAAM T=5) can improve detection accuracy, it generally comes with a higher computational cost, highlighting a trade-off between performance and real-time capability for robotic platforms.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.