Lens: Smart Attendance & Spatial Intelligence Platform

I built Lens as an API-first spatial intelligence engine that converts classroom photographs into verified, deduplicated attendance records:

1. Facial Recognition & 128-d Embedding Pipeline

Utilized Python 3.14, FastAPI, OpenCV, and

dlib

Terminal Workspace
# Facial embedding distance matching with dlib 128-d vectors
def verify_student(detected_encoding, enrolled_encodings, threshold=0.55):
    distances = np.linalg.norm(enrolled_encodings - detected_encoding, axis=1)
    min_idx = np.argmin(distances)
    if distances[min_idx] < threshold:
        return enrolled_encodings[min_idx].student_id, float(distances[min_idx])
    return None, None

2. Multi-Photo Batch Deduplication Engine

Engineered a deduplication queue for multi-angle photo uploads. Lens groups matched faces across frames and selects the single highest-confidence match per student to prevent double-counting.

3. Automated Registrar SIS Sync & Next.js 16 Dashboard

Constructed a visual management dashboard displaying bounding box overlays over lecture photos, confidence meters, and an automated background service syncing attendance directly to university SIS registrar endpoints with full audit logs.