Overview
Facial analysis pipeline estimating age groups and gender from security feed snapshots for retail visitor analytics while preserving privacy.
The Problem
Retail analytics need demographic data of visitors without intrusive interaction or privacy violations.
The Solution
Developed privacy-preserving system that processes frames locally, extracts demographics, and immediately discards images. Only aggregated statistics are stored.
Project Gallery
Results & Impact
±5 years
Age Accuracy
Mean absolute error
90%
Gender Accuracy
On UTKFace benchmark
15 FPS
Processing Speed
Real-time analysis
Key Impact
- Enabled targeted marketing campaigns
- Improved store layout based on visitor demographics
- 100% privacy compliance (no image storage)
Technologies Used
PythonOpenCVKerasTensorFlowMTCNN