This paper presents AgroLens, an offline-first AI system for real-time crop disease detection in low-connectivity agricultural regions.It leverages a lightweight MobileNetV2 CNN model deployed on a low-cost Raspberry Pi 5 device.The model is optimized using pruning and 8-bit quantization to ensure fast and efficient on-device inference.Trained on the PlantVillage dataset, the system achieves 97.45% accuracy with minimal performance loss after compression.The solution eliminates dependency on internet connectivity, enabling accessible and real-time diagnosis for farmers.It demonstrates a scalable edge AI approach for precision agriculture in resource-constrained environments.The system significantly reduces model size and latency while maintaining high diagnostic performance.This work bridges the gap between advanced AI and real-world agricultural deployment.
Kadu et al. (Tue,) studied this question.