1. Enhancing disease region segmentation in rice leaves using modified deep learning architectures.
- Author
-
Sharma, Mayuri, Kumar, Chandan Jyoti, Singh, Thipendra Pal, Talukdar, Jyotismita, Sharma, Rupam Kr, and Ganguly, Ankur
- Subjects
DEEP learning ,RICE quality ,EARLY diagnosis ,RICE ,CROP growth ,CROP yields ,DEEP brain stimulation - Abstract
Rice disease profoundly impacts crop growth and yield. Early disease detection is crucial for effective crop care and treatment. Automated extraction of diseased regions from rice leaves is essential for enhancing automated disease identification systems. In this study, we propose an innovative approach that enhances deep learning (DL) segmentation architecture (UNet) by incorporating dilated convolution, EfficientNetB4, and pixelwise logical AND operation. We focus on three prevalent rice diseases: bacterial leaf blight, brown spot, and leaf smut. Manual ground truth mask images are generated for each disease. A comparative analysis demonstrates the superior performance of the modified architectures over their unmodified counterparts. Notably, the modified UNet model stands out, achieving a mean loss of 0.3018 and a mean dice coefficient of 0.6785 which is a significant improvement compared to conventional UNet model. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF