Research on Mobile Plant Leaf Recognition Based on Lightweight Convolutional Neural Networks

Jingyi Wang

doi:10.6911/WSRJ.202609_12(9).0007

Authors

Jingyi Wang Hebei University of Economics and Business, Shijiazhuang, China

DOI:

https://doi.org/10.6911/WSRJ.202609_12(9).0007

Keywords:

Lightweight Convolutional Neural Network; Plant Leaf Recognition; Mobile Deployment; Model Compression; Deep Learning.

Abstract

For mobile devices, recognizing plant leaves in real time is an important task. But traditional convolutional neural networks are not good at this task. They struggle to work fast enough on these devices. To fix this problem, this paper looks into optimization methods for lightweight convolutional neural networks. The paper picks three baseline models: MobileNetV2, ShuffleNetV2, and EfficientNet-B0. It also uses several compression techniques, such as knowledge distillation, pruning, and quantization. Along with structural improvements, these methods help build lighter models. The paper tests these models on the Flavia dataset. The results are clear. The optimized model reaches a Top-1 accuracy of 97.51%. Its parameter count drops to 1.4M. The model size is reduced by 75%. And its inference speed becomes much faster, which meets the needs of mobile devices.

Downloads

Download data is not yet available.

References

[1] Wang, Z. R., & Yan, C. L. (2011). A review of image feature extraction methods. Journal of Jishou University (Natural Science Edition), 32, 43–47.

[2] Xu, H. T., Cai, W. Y., Zhang, M. Y., & others. (2025). A lightweight convolutional neural network model combining MobileNet and ShuffleNet. Journal of Hangzhou Dianzi University (Natural Sciences), 45, 50–61.

[3] Jin, Z. W., & Ge, D. Y. (2026). Image recognition method based on improved ShuffleNet network. Computer Applications and Software, 43, 182–188.

[4] Zhou, L., & Cao, C. W. (2026). Plant pest and disease recognition method based on knowledge distillation and EfficientNet_v2. Jiangsu Agricultural Sciences, 54, 269–276.

[5] Xiao, Y. (2025). Research on image recognition of crop diseases and insect pests based on lightweight convolutional neural network [Master’s thesis]. University of Science and Technology Liaoning.

[6] Ma, Y., Zhi, M., Yin, Y. J., & Ping, P. (2022). Review of applications of CNN and Transformer in fine-grained image recognition. Computer Engineering and Applications, 58, 53–63.

[7] Molchanov, P., Tyree, S., Karras, T., Aila, T., & Kautz, J. (2017). Pruning convolutional neural networks for resource efficient inference. In International Conference on Learning Representations. https://doi.org/10.48550/arXiv.1608.08710