Lessons from Collecting a Million Biometric Samples

Area of Science:

• Computer Science
• Biometrics
• Machine Learning

Background:

• Independent evaluations are crucial in experimental computer science, particularly for face and gesture recognition.
• Curated datasets are essential for successful evaluations, requiring appropriate design, sufficient size, and comprehensive metadata.

Purpose of the Study:

• To review a ten-year biometric data collection effort.
• To summarize the creation of key biometrics challenge problems.
• To identify challenges and lessons learned from data collection.

Main Methods:

• Review of a decade-long biometric sampling initiative.
• Analysis of data collection designs and execution.
• Identification of challenges encountered during data acquisition.

Main Results:

• Enabled the creation of several key biometrics challenge problems.
• Provided insights into the design and execution of large-scale data collections.
• Documented key challenges and practical lessons learned.

Conclusions:

• A decade of dedicated biometric data collection has facilitated significant advancements in independent evaluations.
• Understanding data collection challenges and lessons learned is vital for developing reliable biometrics datasets and challenge problems.