The Challenge
Traditional manual roll-call attendance across large university lecture halls led to widespread proxy attendance, lost instructional time (averaging 12-15 minutes per lecture), and unreliable LMS reporting. The institution required an instantaneous, contactless biometric solution capable of processing students simultaneously under varying ambient lighting without requiring physical hardware bottlenecks.
Our Solution & Architecture
Elegensoft engineered an edge-computed multi-threaded facial biometric system powered by PyTorch and FastAPI microservices. The model uses lightweight convolutional embeddings cross-referenced against high-dimensional vector embeddings, paired with anti-spoofing liveness detection to completely eliminate photo and screen replay attacks. Real-time sync was integrated directly with the campus SIS and student portals via Flutter mobile applications.
Traditional manual roll-call attendance across large university lecture halls led to widespread proxy attendance, lost instructional time (averaging 12-15 minutes per lecture), and unreliable LMS reporting. The institution required an instantaneous, contactless biometric solution capable of processing students simultaneously under varying ambient lighting without requiring physical hardware bottlenecks.
Elegensoft engineered an edge-computed multi-threaded facial biometric system powered by PyTorch and FastAPI microservices. The model uses lightweight convolutional embeddings cross-referenced against high-dimensional vector embeddings, paired with anti-spoofing liveness detection to completely eliminate photo and screen replay attacks. Real-time sync was integrated directly with the campus SIS and student portals via Flutter mobile applications.