Deep learning with ResNet50 enables flawless cauliflower leaf disease identification, promoting efficient automated solutions for sustainable crop management.
Authors
Sakib Alam Jisan, Daffodil International University, Dept. of Computer Science and Engineering, Dhaka, Bangladesh
Sourav Kumar Das, Daffodil International University, Dept. of Computer Science and Engineering, Dhaka, Bangladesh
Md Julkar Naeen, Daffodil International University, Dept. of Computer Science and Engineering, Dhaka, Bangladesh
Nivedita Haldar, Professor, Jindal Global Business School, O.P. Jindal Global University, Sonipat, Haryana, India
Narayan Ranjan Chakraborty, Daffodil International University, Dept. of Computer Science and Engineering, Dhaka, Bangladesh
Mayen Uddin Mojumdar, Daffodil International University, Dept. of Computer Science and Engineering, Dhaka, Bangladesh
Summary
This study introduces a novel method for cauliflower leaf disease detection using deep learning from raw data. Cauliflower is an economically significant crop that faces substantial challenges due to highly pathogenic leaf diseases, which severely impact yield and quality. The inability to implement effective management practices exacerbates these issues. Accurate disease detection is significant for effective disease management and sustainable agriculture. This study uses a deep learning approach to classify cauliflower leaf diseases, exploring the application of three convolutional neural networks. A dataset comprising 1,598 images was utilized for training and evaluation. Various image pre-processing techniques were applied to enhance model performance. Experimental results indicate that ResNet50 outperformed VGG16 and VGG19, achieving perfect classification accuracy. These findings underscore the precise detection of cauliflower leaf diseases, and the future development of automated agricultural systems.
Published in: Proceedings of the 2025 12th International Conference on Computing for Sustainable Global Development, INDIACom 2025
To read the full article, please click here.
