Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Expanding on the BRIAR Dataset: A Comprehensive Whole Body Biometric Recognition Resource at Extreme Distances and Real-World Scenarios (Collections 1-4)

Gavin Jager, David Cornett, Gavin Glenn, Deniz Aykac, Christi Johnson, Robert Zhang, Ryan Shivers, David Bolme, Laura Davies, Scott Dolvin, Nell Barber, Joel Brogan, Nick Burchfield, Carl Dukes, Andrew Duncan, Regina Ferrell, Austin Garrett, Jim Goddard, Jairus Hines, Bart Murphy, Sean Pharris, Brandon Stockwell, Leanne Thompson, Matthew Yohe

TL;DR

The paper addresses biometric identification at extreme distances and elevated viewpoints by expanding the BRIAR dataset with Collections 3 and 4 (BGC3/BGC4), incorporating more locations, sensors, and realistic scenarios including group activities and a Hogan's Alley mock-city. It presents a comprehensive data collection, curation, and annotation pipeline, along with an evaluation protocol design (BRS/BTS, FaceIncluded vs FaceRestricted, Simple vs Blended galleries) to benchmark whole-body recognition under challenging conditions. Key contributions include substantial dataset growth (over 475k images, 3,450 hours of video from 1,760 subjects), a robust curation and QA workflow, automated and manual annotations, and detailed privacy protections and IRB governance. The work enables researchers to develop more robust, generalizable biometrics systems for security-relevant tasks across long ranges, elevated sensors, and realistic operational contexts, with implications for equitable performance and deployment readiness.

Abstract

The state-of-the-art in biometric recognition algorithms and operational systems has advanced quickly in recent years providing high accuracy and robustness in more challenging collection environments and consumer applications. However, the technology still suffers greatly when applied to non-conventional settings such as those seen when performing identification at extreme distances or from elevated cameras on buildings or mounted to UAVs. This paper summarizes an extension to the largest dataset currently focused on addressing these operational challenges, and describes its composition as well as methodologies of collection, curation, and annotation.

Expanding on the BRIAR Dataset: A Comprehensive Whole Body Biometric Recognition Resource at Extreme Distances and Real-World Scenarios (Collections 1-4)

TL;DR

The paper addresses biometric identification at extreme distances and elevated viewpoints by expanding the BRIAR dataset with Collections 3 and 4 (BGC3/BGC4), incorporating more locations, sensors, and realistic scenarios including group activities and a Hogan's Alley mock-city. It presents a comprehensive data collection, curation, and annotation pipeline, along with an evaluation protocol design (BRS/BTS, FaceIncluded vs FaceRestricted, Simple vs Blended galleries) to benchmark whole-body recognition under challenging conditions. Key contributions include substantial dataset growth (over 475k images, 3,450 hours of video from 1,760 subjects), a robust curation and QA workflow, automated and manual annotations, and detailed privacy protections and IRB governance. The work enables researchers to develop more robust, generalizable biometrics systems for security-relevant tasks across long ranges, elevated sensors, and realistic operational contexts, with implications for equitable performance and deployment readiness.

Abstract

The state-of-the-art in biometric recognition algorithms and operational systems has advanced quickly in recent years providing high accuracy and robustness in more challenging collection environments and consumer applications. However, the technology still suffers greatly when applied to non-conventional settings such as those seen when performing identification at extreme distances or from elevated cameras on buildings or mounted to UAVs. This paper summarizes an extension to the largest dataset currently focused on addressing these operational challenges, and describes its composition as well as methodologies of collection, curation, and annotation.
Paper Structure (20 sections, 11 figures, 2 tables)

This paper contains 20 sections, 11 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. Contributions
  3. BACKGROUND AND RELATED WORK
  4. DATA COLLECTION METHODS
  5. Privacy, Security, and Well-being
  6. Controlled Collection
  7. Field Collection
  8. BGC4 Mock City: Hogan's Alley
  9. DATASET SUMMARY
  10. DATA CURATION METHODOLOGY
  11. Data Curation Pipeline
  12. Data Cleaning and Compilation
  13. Video and Image Extraction
  14. XML Metadata Generation
  15. Quality Assurance and Finalization
  16. ...and 5 more sections

Figures (11)

  • Figure 1: Sample images from the BRIAR dataset. Subjects in figures have consented to appearing in publications.
  • Figure 2: Sample frames of videos captured by various platforms during BGC3 and BGC4 (clockwise from the top-left): ground camera at 500m, rooftop camera at 720m, elevated close-range camera, UAV platform.
  • Figure 3: Frames of subjects participating in some of the new field scenarios (clockwise, starting from the top-left): "backpack", "cell phone", "box stack", and "group backpack".
  • Figure 4: Overview of the BGC3 field collection layout. Labelled areas: (i) close-range mast-mounted cameras, (ii) fish-eye weather camera, (iii) outdoor collection tent, (iv) UAV control center, (v) UAV landing zone, (vi) 100-m range cameras, (vii) 200-m range cameras on scaffolding, (viii) 300-m range cameras, (ix) 380-m range cameras, (x) 400-m range cameras, (xi) 500-m range cameras
  • Figure 5: Overview of the BGC4 field collection layout. Labelled areas: (i) close-range mast-mounted cameras, (ii) fisheye weather camera, (iii) outdoor collection tent, (iv) UAV control center, (v) UAV landing zone, (vi) 180-m range cameras, (vii) 330-m range rooftop cameras, (viii) 140-m range camera, (ix) 720-m range rooftop cameras
  • ...and 6 more figures