| dc.contributor.advisor | Tomato is highly grown vegetable all over the
world. Tomato is highly susceptible to diseases and
considerable amount of crop is wasted due to diseases
caused by virus, bacteria and fungi. Disease identification
of Tomato is a major problem faced by farmers. The
proposed system helps farmers to identify four tomato
diseases namely Anthracnose, Blossoms End Rot, Late
Blight and Powdery Mildew. Convolutional Neural
Network (CNN) has been applied in the study to predict the
disease from the images. The implementation of CNN from
scratch demands high computational resources and
considerable amount of image data. Therefore, transfer
learning approach has been applied with the MobileNet
model which is trained on the ImageNet classification
dataset. This research work was conducted by changing
the number of images, training models and
hyperparameters to experiment the accuracy of the
system. The system gained 99.16% of training accuracy,
98.89% of validation accuracy and 98.96% of test accuracy
with 0.0001 learning rate, 0.9 momentum, batch size as 32
and 3200 training images. | |
