Snow Leopard Identification

Computer vision system for identifying individual snow leopards using feature matching and machine learning, supporting conservation efforts in Central Asia.
🧑🏾‍🏫 Manual 📖
  1. Upload Snow Leopard Image: Select an image of a snow leopard showing distinctive coat patterns. The system works with multiple body parts including head, flanks, and tail.
  2. Configure Matching: Optionally filter by location and body part, then click 'Submit' to process the image.
  3. Review Matches: The system displays the top matching individuals from the catalog with confidence scores. High confidence matches (🔵 Excellent, 🟢 Good) indicate likely the same individual.
  • 🛠️ Learn more about the project here.
  • The code is available here.
  • 🤗 The application is hosted here.

Overview

The Snow Leopard Identification system enables researchers and conservationists to track individual snow leopards across Central Asia’s mountain ranges without invasive tagging. By analyzing the unique rosette patterns on each animal’s coat, the system creates distinctive visual fingerprints that remain stable throughout the animal’s life.

Snow leopards are notoriously difficult to study due to their remote, high-altitude habitats spanning rugged terrain across 12 countries. This technology transforms camera trap images into actionable conservation data, supporting:

  • Individual identification across survey seasons
  • Population size estimation in remote regions
  • Movement tracking between territories
  • Long-term survival monitoring

Key Features

  • Multi-Body-Part Matching: Works with images of head, left flank, right flank, tail, and other body regions
  • Location Filtering: Filter catalog searches by geographic region
  • Multiple Feature Extractors: Choose between SIFT, SuperPoint, DISK, or ALIKED for optimal matching
  • Confidence Scoring: Four-level confidence system (🔵 Excellent, 🟢 Good, 🟡 Fair, 🔴 Uncertain)
  • Visual Verification: Side-by-side comparison views with matched keypoint visualization

Use Cases

  • Population Monitoring: Estimate population sizes in protected areas by identifying unique individuals from camera trap surveys. Track how populations change over time without physical capture.
  • Movement Ecology: Understand how individual snow leopards move across landscapes by matching sightings at different camera trap locations. Map territory boundaries and corridor usage.
  • Conservation Planning: Identify key individuals and breeding populations to prioritize protection efforts. Monitor the effectiveness of conservation interventions.

Learn More

For complete details on the snow leopard monitoring project and partnership with OSI Panthera, visit the project page.