DeepPerson

A simple person re-identification system with automatic person detection and embedding generation.

Installation

pip install deep-person

Quick Start

from deep_person import DeepPerson

dp = DeepPerson()

# Generate embeddings
result = dp.represent("person.jpg")
print(f"Total subjects detected: {len(result.subjects)}")

# Verify if two images show the same person
result = dp.verify("person1.jpg", "person2.jpg")
print(f"Same person: {result.verified}")

# Batch processing - multiple images at once
results = dp.represent(["person1.jpg", "person2.jpg", "person3.jpg"])
print(f"Total subjects detected: {len(results.subjects)}")

Features

• Automatic person detection - YOLO-based detection
• Multi-modal embeddings - Body (2048-dim) + Face (512-dim) or via Model Registry
• Identity verification - Cosine/Euclidean distance metrics
• GPU acceleration - Automatic CUDA detection with CPU fallback
• Batch processing - Efficient multi-image handling

Requirements

• Python 3.11+
• Models downloaded automatically on first use
• Optional: CUDA GPU for faster processing