Abstract:
The king of fruits, the mango is in great demand, thus it is imperative to prevent sickness
in order to maximize profits. Because symptoms can vary, automatic segmentation and
detection of leaf diseases remains difficult. For any computer-aided method to identify leaf
diseases in mango plants, accurate segmentation of the illness is a critical requirement. I
suggested using a CNN model to partition the mango leaf's sick area in order to address
this problem. The suggested CNN uses certain preprocessing methods before learning the
characteristics of each pixel in the input data directly. Using the real-time dataset provided
by the Bangladeshi mango research, we assessed the suggested CNN. I compared the
segmentation performance of the suggested model results to that of the state-of-the-art
models that are currently available, such as Vgg16, resnet50, denseNet, efficientNet,
InceptionV3, and mobileNet. Moreover, the segmentation accuracy of the suggested model
training is 99.46%, significantly greater than that of the other models. We have determined
that by utilizing a CNN, the input image might learn more distinct and focused features,
leading to an enhanced recognition performance and the identification of disorders. As a
result, I also make a web application. given that most individuals in our country know how
to use an Android feature phone. I write a simple software. Here, all the user needs to do
is snap a photo of an mango leaf, and my algorithm will tell you whether it's healthy or not.
