|
|
 |
| |
|
|
|
Aiding the Digital Eyes to Detect the Leaf-Spot Diseases in Rice Crops using DenseNet CNN Algorithms |
|
|
|
PP: 1253-1262 |
|
|
|
doi:10.18576/amis/190602
|
|
|
|
Author(s) |
|
|
|
Suleiman Shelash Mohammad,
G. Jegan,
N. Raja,
Hanan Jadallah,
M. A. Muthiah,
P. Kavipriya,
R. M. Joany,
T. Vino,
Asokan Vasudevan,
Zhidong Feng,
|
|
|
|
Abstract |
|
|
| To investigate the capability of deep learning models in identifying rice crop diseases. A convolutional neural network (CNN) based classification strategy was introduced in this re-search for high-resolution agricultural remote sensing pictures of paddy leaves. Various clas-sifications were tested and assessed based on specificity, accuracy, recall, precision, and F1-score. To analyse and comprehend the technique, a substantial quantity of samples obtained from open source panchromatic pictures with a resolution of 256x256 pixels were used for experimentation and verifying using MATLAB. This research relies on datasets that were acquired from the Kaggle repository. The dataset used for rice disease identification consists of 3,400 photos. These images are divided into 60 images for each of the seven different dis-ease categories such as Bacterial Blight, Downy mildew, brown spot, Dead Heart, Hispa, leaf blast, Tungro. There were a total of 550 photographs that were specifically chosen for valida-tion. The performance of DenseNet 201 under CNN was comparatively higher with the accuracy of 97% and precision of 99%.Later the study was extended towards the modified DenseNet, where in accuracy and elapsed time where improved eliminating the traditional three convolutional layers of DenseNet 201. The study also incorporated an extension to the modified DenseNet architecture, leading to improved accuracy and decreased elapsed time. An improvement was made by removing the traditional three convolutional layers found in DenseNet 201. |
|
|
|
|
|
|
|
