Dataset Card for Event Human Motion Tracking Dataset

Table of Contents

• Dataset Description
  • Dataset Summary
  • Supported Tasks
• Dataset Structure
  • Directory Layout
  • File Descriptions
• Dataset Creation
  • Sensors & Hardware Setup
  • Data Collection & Scenarios
• Usage & Integration
• Citation

Dataset Description

Dataset Summary

The Event Human Motion Tracking Dataset is a dataset designed for fine-grained human motion detection, tracking, and gait/pose analysis. Recorded with a 76° Field of View (FoV) event camera setup, the dataset explicitly evaluates the performance impact of Default vs. Optimized sensor bias parameters on background noise, visual clarity, and object tracking robustness.

The dataset provides synchronized .avi video recordings for visual inspection, algorithmic benchmarking, and spatial ground-truth verification across both bias configurations.

Supported Tasks

• Human Motion Tracking: Bounding box detection and temporal association across dynamic motion recordings.
• Sensor Bias Benchmarking: Comparative evaluation of default vs. custom-optimized sensor bias settings ($76^\circ$ FoV).
• Neuromorphic Motion Analysis: Feature extraction, velocity estimation, and performance assessment under varied bias regimes.

Dataset Structure

Directory Layout

The repository contains .avi video recordings separated by bias configuration, alongside compressed archives for annotations and parameter settings:

Event-Human-Motion-Tracking-Dataset/
├── Default-Bias-Data-76-Degree-FoV/     # Video recordings captured with Default Bias (.avi)
│   ├── recording_01.avi
│   └── recording_02.avi
├── Optimized-Bias-Data-76-Degree-FoV/   # Video recordings captured with Optimized Bias (.avi)
│   ├── recording_01.avi
│   └── recording_02.avi
├── Bias-Settings-Real-Time.zip          # Real-time sensor bias configuration logs and register settings
├── Mixed-Bias-Data-Annotations.zip      # Ground truth bounding box annotations and track IDs
└── README.md