Abstract:
Our farmers provide food for the entire planet. However, they face a wide range of
obstaclesin agriculture, which they are unable to manage, including from illnesses to insect
infestations. Crop losses are caused by a number of diseases, and diseases are a part of our
normal life. Using a computer based automatic system or approach, like image
segmentation for disease recognition systems or approaches. It is really beneficial to us.
For this most worthwhile apparatus is deep learning in convolution neural network CNN.
This paper proposes some methodology to detect 3 types of rice diseases with U-Net
architecture. We worked on pre-processed data with three trained models(60:40; 70:30;
80:20). In our research we segmented RGB image to grayscale output using semantic
segmentation. We also used the Grid Search algorithm for comparing six types of
optimizer. We used Adam optimizer to run this model. We developed a U-Net architecture
model with added layers and successfully got more than 93% accuracy respectively which
is more efficient for future deep learning and also the agricultural sector.
