Vol. 11, Issue 7 (2022)

Author(s):
N Vinoda, Premkumar Borugadda, Vimala Beera and Ravi Babu M

Abstract:
In Precession agriculture, computer vision has been demonstrated as state-of-the-art technology. In this paper, a VGG16 model was applied to identify and classify cotton leaf diseases. Cotton Dataset consists of 2204 images, in which 1951 images were used for training and 253 images were used for validation. Apply the transfer learning on thirteen convolutional layers of VGG16 for extracting the features on 1951 images. 25088 features are extracted by transfer learning. With these features form high dimensions, if we apply any classification algorithms on high dimension model, may get over fitted. So, for reducing large dimensions, use one dimension reduction technique, namely Principal component analysis (PCA). The output of PCA is low dimension. Now apply three fully connected layers of VGG16 and machine learning classification algorithms on low dimension data. Three fully connected layers of VGG16 provided the best performance model with a 95.65% validation accuracy at the training time of about 140 seconds.