Hi,

First define what will be the exact inputs and outputs of your system. Inputs will be IMAGES? What size/resolution images? Full/Partial leaf image? Full/Partial plant image? Blurry/Sharp/Bright images? How are the diseases detected by humans? Based on change in colour of full/partial leaf? Change in Shape? How many types of leaves? How many types of diseases?

Once all this is clear...Then most probably you are looking for Image classification...

Image classification means you take input an image. And assign each disease category into a class. Train the network and and then you can use the neural network for testing. The trained network will take input an image and will tell you which of the output CLASSES/DISEASES it belongs to.

You can use computer vision algorithms and relevant leaf databases. For computation you may use Matlab, e.g., discussion: https://www.mathworks.com/matlabcentral/fileexchange/55098-plant-leaf-disease-detection-and-classification-using-multiclass-svm-classifier

A good Pattern recognition is attached.

I would recommend you to first threshold the image using Otsu method.You can then use Morphological snake algorithm to detect anomalies in the contours formed in the image of the leaf. If you couple it with machine learning,you will get a more efficient result.

@Yasir Jan I have a labelled dataset from kaggle that I will be using and yes It is a classification problem.

@Muhammad Ali and @Anuska Kundu thanks for your reply.

Of course, this is a classification problem. Nosological images are classes. The spectral characteristics of individual leaves can be used as attributes

Article Intellectual analysis of spectral images of plants

@Eugene Veniaminovich Lutsenko thanks for the article.

Firstly, you need to understand what features or data you have about the plant(s) which you can apply/use for the disease prediction. In other words, prior knowledge of the available features/representations of the plant(s) will enable you to decide on whether your Machine Learning approach to the problem will be based on either Supervised Learning, Unsupervised Learning, or Semi-Supervised Learning.

Thereafter, you can choose to either model your disease prediction model as either a Regression model (prediction via degree/percentage of accuracy or fitness) or Classification model (prediction via feature classification).

You can try this method deep stacked varitional neural network

here its python code... and the details in the paper ..

Code Python code Deep Stacked Varitional Neural Network

Conference Paper Industrial Internet of Things Based Ransomware Detection usi...