Bharat ABIS: Billion-Scale Multi-Modal Biometric Search
A new paper introduces Bharat ABIS, the first open-source system for billion-scale multi-modal biometric search, designed for country-level identity systems. The system processes fingerprints, face, and iris modalities through a pipeline of preprocessing, quality assessment, presentation attack detection, and feature extraction, producing a 13.5KB template per person. It addresses challenges such as acquisition, matching speed, and handling missing finger digits. The paper provides detailed analysis of modality integration for efficient 1:N search (de-duplication).
Key facts
- First paper on billion-scale multi-modal biometric search
- System called Bharat ABIS
- Based on open-source architectures
- Processes fingerprint, face, and iris modalities
- Produces a concatenated template of 13.5KB per person
- Designed for country-level identity systems
- Addresses acquisition, preprocessing, feature extraction, accuracy, matching speed, presentation attack detection, and special cases like missing finger digits
- Provides detailed analysis of modality integration for 1:N search
Entities
—