https://doi.org/10.1140/epjs/s11734-025-01700-1
Regular Article
EPC-GANet: a lightweight attention guided network with expanded receptive field for rice leaf disease recognition
College of Science, Northeast Forestry University, Hexing Road 26, 150040, Harbin, Heilongjiang, People’s Republic of China
Received:
21
March
2025
Accepted:
19
May
2025
Published online:
5
June
2025
Rice is among the most significant staple crops for humanity, and its yield frequently suffers substantial losses due to various diseases. To address the detection delay issues caused by the high parameter count, reliance on cloud services, and inadequate network coverage in traditional models, a novel model solution named enhanced partial convolution guided attention network (EPC-GANet) was proposed. Under the premise of ensuring detection accuracy, it achieves model miniaturization and real-time detection, breaking the technical bottleneck of local mobile device deployment for rice leaf disease real-time detection. By integrating the partial convolution downsampling extension module (PCDEM) and the enhanced guidance attention mechanism (EGAM), EPC-GANet could effectively focus on the significant features of rice leaf diseases. Additionally, by introducing the MaxDepth Pooling module in the shallow structure of the model, its stability was further enhanced. Experimental results showed that the EPC-GANetV1 model required only 29.22 min for training on 10,788 sample images, with a model size of just 0.97 MB, making it suitable for scenarios with high real-time requirements. EPC-GANetV4 achieved an accuracy rate of 97.12% while maintaining its lightweight characteristics, making it more suitable for scenarios with high-precision requirements. Based on the EPC-GANetV4 model, a rice leaf disease diagnosis application that integrated real-time detection and image acquisition functions was developed, achieving offline real-time diagnosis in agricultural scenarios. This study significantly improves the environmental adaptability and operational stability of the disease control system by introducing a guided attention model, providing strong support for the development of smart agriculture.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
