TFH - Face Deduplication Collection

We are collecting a large and diverse dataset of current neutral selfies, head-pose captures, and historical facial images to support machine-learning research and facial recognition model training. The focus is on capturing real-world variation across lighting, poses, expressions, accessories, environments, and aging to improve model accuracy and robustness.

About the role

• Current Neutral Selfies – clean frontal selfies serving as high-quality identity references, with natural variation in appearance and surroundings.
• Current Head-Pose Captures – selfies captured in assigned head directions to introduce pose variation.
• Historical Images – older photos from participants’ galleries to capture natural aging and long-term appearance changes.

Requirements

• Submit a minimum of 20 valid images to qualify for payment.
• Maximum payout is based on 24 accepted images.
• Due to strict automated and manual Quality Control (QC), we recommend submitting 30 images to ensure enough remain valid after review.
• Use a Gmail address as your primary account when applying for this project.

Pay

Compensation rate is $0.55 USD per accepted image.