app/Hackathon_setup/README.md

metadata

title: Face Recognition System
emoji: 👤
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.0.0
app_file: app.py
pinned: false
license: mit
short_description: A face recognition system using Siamese networks and cosine similarity

Face Recognition System

A comprehensive face recognition system that can compare faces for similarity and classify faces into different categories.

Features

• Face Similarity Comparison: Compare two face images and get a similarity score
• Face Classification: Classify a single face image into predefined categories
• Cosine Similarity: Uses advanced cosine similarity for accurate face matching
• Siamese Neural Network: Leverages deep learning for robust feature extraction

How to Use

Face Similarity

1. Upload two face images
2. Click "Compare Faces"
3. Get a similarity score and interpretation

Face Classification

1. Upload a single face image
2. Click "Classify Face"
3. Get the predicted face class

Similarity Score Interpretation

• 0.8-1.0: Very High Similarity (likely same person)
• 0.6-0.8: High Similarity (possibly same person)
• 0.4-0.6: Moderate Similarity (uncertain)
• 0.2-0.4: Low Similarity (likely different persons)
• 0.0-0.2: Very Low Similarity (definitely different persons)

Technical Details

• Model: Siamese Neural Network with CNN layers
• Similarity Metric: Cosine Similarity
• Classifier: K-Nearest Neighbors
• Preprocessing: StandardScaler normalization
• Input Size: 100x100 pixels (grayscale)

API Usage

The system also provides API endpoints:

Similarity API

curl -X POST \
  -F "file1=@face1.jpg" \
  -F "file2=@face2.jpg" \
  https://pavaniyerra-hackthon4.hf.space/predict_similarity/

Classification API

curl -X POST \
  -F "file=@face.jpg" \
  https://pavaniyerra-hackthon4.hf.space/predict_class/

Requirements

• Python 3.11
• PyTorch
• OpenCV
• scikit-learn==1.6.1
• Gradio
• NumPy
• Pillow
• Joblib

Model Files

The system requires these trained model files:

• siamese_model.t7: Trained Siamese network
• decision_tree_model.sav: KNN classifier
• face_recognition_scaler.sav: StandardScaler for normalization