Three-Dimensional Convolutional Neural Network for Wheat Rust Disease Classification

In worldwide, the wheat is a significant crop which generates main source of food for numerous peoples. Through, the development of crop productions is vulnerable through various diseases such as bacterial, viral and fungal infections. These type of disease can cause important damage in crops which leads to diminish the yield production and grain quality. This paper proposed a Three-Dimensional Convolutional Neural Network (3D-CNN) for wheat rust disease classification that learns to recognize the pattern and structure of rust disease using convolution filter layers. The CGIAR dataset is used which contains 1486 images and it is pre-processed by gaussian filter which reduces the noise and smoothens the image. Then, the Discrete Wavelet Transform (DWT) is used for feature extraction which works in discrete timespan that outputs in low computational cost. Then, 3D-CNN is used for the classification of wheat rust disease. The performance of 3D-CNN is estimated by accuracy, recall, f1score and precision. The 3D-CNN attains accuracy 99.83%, recall 98.89%, f1score 98.81 % and precision 98.79
^%
when compared to existing techniques like GNet+FERSPNET-50 and Few-shot learning based EfficientNet.