Deep Learning for Single-Photon Emitter Image Classification

The patent replaces the mechanical/optical HBT experiment system with a computational deep learning system. Instead of performing physical correlation measurements through complex optical paths and timing electronics, the system uses trained neural network models to classify emitters based on image data, substituting a time-consuming physical measurement process with a computationally efficient classification approach

The patent creates a computational model (deep learning network) that learns from training data to replicate and generalize the identification capability of HBT measurements. The model captures the essential characteristics of single-photon emitters from training examples and can quickly identify them in new data without requiring actual HBT measurements, effectively copying the measurement function through learning rather than direct measurement

The patent replaces manual or semi-manual HBT experiment operations with an automated deep learning classification system. The system automatically processes images, extracts features, and classifies emitters without requiring skilled operators to perform complex correlation measurements, thereby reducing labor intensity while maintaining determination accuracy

The deep learning system performs self-service by automatically learning from training data and independently classifying new emitter data without requiring human intervention in the measurement or analysis process. The model autonomously identifies single-photon emitters based on patterns learned during training, eliminating the need for manual operation of HBT experiments