Facial Recognition/Verification Data Collection Project
Shaip | Powering Ethical AI with Quality Data
Introduction
Welcome to Shaip’s Facial Recognition/Verification Data Collection Project, conducted through the Shaip App. Your participation will help develop highly accurate and inclusive biometric systems for real-world applications.
This project focuses on gathering diverse facial data to train AI-based recognition technologies that work across varied environments, expressions, and accessories—minimizing bias and improving system reliability.
Project Objective
To collect high-resolution facial images and videos with diverse expressions, movements, and environmental conditions to train and validate next-generation facial recognition AI models.
Please fill the details below & we’ll be in touch
Fill out the form or reach us on: workforce@shaip.in
Who Can Participate?
Criteria | Details | Eligibility Criteria |
---|---|---|
Gender | Female Only |
Participants must be able to submit 5 historical selfie images taken over the past 12 months in varying lighting conditions (morning, evening, night). If you cannot provide these images, please do not apply. |
Skin Tone | White | |
Age | 30 years and above | |
Total Participants Required | 400 women |
Data Collection Requirements
Type | Deliverables Details |
---|---|
Videos | 4 short videos (5–10 seconds each) capturing head movements, expressions, and accessories |
Images | 1 live selfie + 5 historical selfie images from the past year |
Task-Wise Recording Guidelines
Task | Description | Head Movements | Facial Expression | Lighting & Environment | Duration |
---|---|---|---|---|---|
Task 1 | Natural Head Movement (No Accessories) Perform controlled head rotations to capture natural movement without expression changes. |
1. Yaw (Side-to-Side) – Slowly turn your head left and right. 2. Pitch (Up-and-Down) – Nod your head up and down. 3. Roll (Side Tilt) – Tilt your head towards each shoulder. |
Neutral | Use Low, Moderate, and Bright lighting over different attempts if possible. Specify location: Indoor/Outdoor |
5–10 seconds |
Task 2 | Smiling Expression Repeat Task 1 movements while smiling. |
Smiling | Moderate or Bright lighting only. Specify location: Indoor/Outdoor |
5–10 seconds | |
Task 3 | Frowning Expression Repeat Task 1 movements while frowning. |
Frowning | Moderate or Bright lighting only. Specify location: Indoor/Outdoor |
5–10 seconds | |
Task 4 | Accessories & Makeup Variations Wear accessories (hat, scarf, glasses, face jewelry, makeup on/off, different hairstyles). Perform the same movements as Task 1. |
Neutral | Moderate or Bright lighting only. Ensure accessories don’t block the face. Specify location: Indoor/Outdoor |
5–10 seconds |
Lighting Guidelines
Lighting Level | Lux Level | Example |
---|---|---|
Low Light | < 450 lux | Dim room with one small light |
Moderate Light | 450–750 lux | Standard indoor lighting (LED bulbs, window light) |
Bright Light | > 750 lux | Outdoor in sunlight or under bright studio lights |
Image Collection Details
Image Type | Description |
---|---|
Live Selfie | One current selfie showing the face and top of shoulders. |
5 Past Selfies | Selfies taken within the past year, in different lighting conditions (morning, evening, night). Expressions may vary. |
Recording & Device Instructions
Device Requirements
- Use a smartphone (Android or iOS).
- Provide your phone make and model when submitting.
Recording Quality
Parameter | Requirement |
---|---|
Resolution | 1080p (Full HD) |
Frame Rate | 30 fps |
Orientation | Portrait (Vertical) |
Focus | Face must be sharp and clear at all times |
Background | Neutral, clutter-free, no other people or moving objects |
Stability | Use a steady hand or tripod. Avoid shaky footage. |
Technical Setup Guide
Device | Steps |
---|---|
For Android Users | 1. Open Camera Settings. 2. Select Video Resolution: 1080p (Full HD). 3. Set Frame Rate: 30 fps. 4. Record a test video to verify clarity. |
For iOS Users | 1. Go to Settings → Camera → Record Video. 2. Choose 1080p at 30fps. 3. Record a test video to check quality. |