Fast and Robust Analysis for Quantum Sensing Spectra via Deep Learning

Nitrogen-vacancy (NV) centers in diamond are premier sensors for nanoscale measurements of magnetic fields, temperature, and strain. However, conventional analysis of ODMR spectra via nonlinear least-squares fitting is computationally expensive, sensitive to initialization, and unreliable at low signal-to-noise ratios (SNR). Here, we introduce a machine-learning framework based on a one-dimensional convolutional neural network (1D-CNN) that directly maps ODMR spectra to resonance parameters without iterative optimization. This approach enables highly parallel GPU execution for real-time analysis. Benchmarking against synthetic and experimental datasets demonstrates that our model outperforms standard fitting in accuracy and noise tolerance, particularly in low-SNR regimes. We validate the framework through two distinct applications: widefield imaging of magnetic vortices in high-$T_c$ superconductors and intracellular thermometry using nanodiamonds. These results establish ML-based inference as a robust, high-throughput solution for complex quantum sensing data analysis.