Answers are grounded in this paper and cite their sources.
The need for more efficient early stress monitoring systems in smart agriculture is becoming increasingly imperative due to the rising frequency of natural disasters. Nanosensors offer a solution for non-destructive acquisition of high-throughput plant molecular signals. However, mining these time-series data, which vary across different time scales, lacks updated definitions and well-designed neural network modules. Here, we developed a strategy for early stress prediction in nano-smart plants (DeepSPN) based on fluorescent nanosensors and deep learning (DL). This system can rapidly identify and classify in-situ signals, enabling remote monitoring of plant health via the Web from various mobile devices. In this system, surface-modified single-walled carbon nanotubes (SWCNTs), which emit in the near-infrared (NIR) region, were employed to construct specific molecular recognition sites for stress-induced H₂O₂ in leaves, thereby enabling species-independent analysis of signaling kinetics. Owing to an attention mechanism-based multi-scale convolutional neural network (AMCN) and a risk-sharing strategy, DeepSPN achieved over 80% accuracy with test samples as short as 30 minutes. Based on this strategy, we built a low-cost Raspberry Pi–based wireless solution that achieves detection performance comparable to InGaAs detectors, providing a feasible reference for field monitoring.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: