Problem
Industrial defect classification needs a consistent way to compare classical and hybrid quantum-classical inference under clean and noisy inputs.
Quantum-Hybrid Defect Detector
A full-stack industrial defect-classification application comparing classical CNN and hybrid quantum-classical inference paths.
Architecture
Vue 3 frontend with bilingual UI and dashboards calls a FastAPI backend. The backend loads CNN, QNN_CPU, and CUDA-conditional QNN_GPU models and exposes classification, benchmark, and quantum-advantage routes.
Role
Programming lead for the hybrid-fusion system, stateless backend, model warm-up, rate limiting, secure headers, and Vue SPA.
Results
On 2,833 test images, CNN reached 93.47% accuracy and 93.70% macro F1, QNN_CPU reached 94.88% and 94.99%, and QNN_GPU led with 95.59% and 95.67%.