Efficient and precise quantum state detection is essential for quantum information processing, serving as the backbone for high-fidelity initialization, real-time error correction, and scalable quantum operations. In particular, fluorescence-based detection, which is widely used in both neutral-atom and trapped-ion systems, must retain high fidelity under varying imaging conditions. Here, we present an adaptive convolutional neural network (CNN) that incorporates exposure time as a contextual input, enabling robust generalization across different exposure settings without retraining. To meet the stringent requirements of real-time quantum applications, we accelerate the model using a field-programmable gate array (FPGA), achieving an inference latency of $9.85\text{μ}\mathrm{s}$ at a 500 MHz clock frequency. Compared with state-of-the-art CPU and graphics processing unit (GPU) implementations, our hardware-accelerated design achieves an inference latency reduction of about $30\times$. These results highlight the potential for real-time context-aware quantum state detection, with a scalable model design that can incorporate additional experimental parameters. The proposed approach is well suited for integration in heterogeneous control architectures, supporting low-latency, resource-efficient, hybrid quantum-classical workflows.

Close

PrevNext

View figure in article

• Convolutional neural networks
• Quantum circuits
• Quantum computation
• Quantum information with atoms & light
• Quantum information with trapped ions
• Qubits
• Rydberg atoms & molecules
• Fluorescence