发表机构
University of Science, VNU-HCM; Vietnam National University, Ho Chi Minh City; Van Lang University（越南国立大学胡志明市理科大学; 越南国立大学胡志明市分校; 文朗大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出仅193K参数的轻量级网络TinyCNN用于设备端植物病害检测，在PlantVillage上达98.88%准确率，并通过跨数据集诊断揭示鲁棒性差距。
AI 中文摘要
早期检测作物病害对于联合国可持续发展目标2（零饥饿）下的可持续农业和粮食安全至关重要，在资源受限地区尤为紧迫，这些地区专家诊断稀缺但低成本移动设备普及。本文提出TinyCNN，一种用于设备端植物病害分类的轻量级卷积神经网络。TinyCNN采用深度可分离卷积块，仅含193,190个可训练参数，对于224x224输入图像计算量为110.05M MACs。在38类PlantVillage基准上，TinyCNN达到98.88%的测试准确率和98.03%的宏F1分数，同时比ResNet18小约58倍，比MobileNetV2教师模型小11.8倍，直接降低了推理的能耗、内存和成本足迹，符合绿色AI原则。本文进一步分析普通知识蒸馏作为可持续模型压缩策略；对alpha在{0.3, 0.5, 0.7}和T在{2, 4}中的消融实验选择alpha=0.3, T=4，产生蒸馏后的TinyCNN，测试准确率为98.81%。最后，从PlantVillage到PlantDoc的跨数据集评估揭示了在真实世界条件下的显著鲁棒性差距，Grad-CAM分析将其归因于叶外、背景驱动的注意力，与捷径学习一致。因此，TinyCNN是可持续农业智能的节能、可部署构建模块，而现场鲁棒性仍是实现持久现实影响的关键障碍。
英文摘要
Detecting crop disease early is central to sustainable agriculture and food security under United Nations Sustainable Development Goal 2 (Zero Hunger), and is especially urgent in resource-constrained regions where expert diagnosis is scarce but low-cost mobile devices are widespread. This paper presents TinyCNN, a lightweight convolutional neural network for on-device plant disease classification. TinyCNN uses depthwise separable convolution blocks and contains only 193,190 trainable parameters with 110.05M MACs for a 224x224 input image. On the 38-class PlantVillage benchmark, TinyCNN achieves 98.88% test accuracy and 98.03% macro-F1 while being approximately 58x smaller than ResNet18 and 11.8x smaller than a MobileNetV2 teacher, directly reducing the energy, memory, and cost footprint of inference in line with Green AI principles. The paper further analyzes vanilla knowledge distillation as a sustainable model-compression strategy; an ablation over alpha in {0.3, 0.5, 0.7} and T in {2, 4} selects alpha=0.3, T=4, producing a distilled TinyCNN with 98.81% test accuracy. Finally, cross-dataset evaluation from PlantVillage to PlantDoc reveals a substantial robustness gap under real-world conditions, which a Grad-CAM analysis attributes to off-leaf, background-driven attention consistent with shortcut learning. TinyCNN is thus an energy-efficient, deployable building block for sustainable agricultural intelligence, while field robustness remains the key barrier to durable real-world impact.
Comments13 pages, 3 figures. Accepted at ISRSD 2026
