Back to Projects

Project #004

Face Recognition and Liveness Detection API

Face anti-spoofing system with custom-trained model

Overview

A Python-based API for face recognition and spoofing detection. Trained a custom liveness detection model using MobileNetV2 on a Kaggle dataset, achieving over 80% validation accuracy. The model was exported to ONNX for production use.

Key Features

  • Custom-trained liveness detection model
  • Face recognition and spoofing prevention
  • Model export to ONNX format for efficient deployment
  • Real-time validation with 80%+ accuracy

Challenges & Solutions

Training an accurate liveness model required extensive tuning and early stopping techniques. After multiple iterations, achieved a stable validation accuracy of 80%, ensuring robust spoofing detection.

Technologies

PythonPyTorchFastAPIONNXMobileNetV2