# An Effective Approach for Recognition of Crop Diseases Using Advanced Image Processing and YOLOv8

## TL;DR

This paper introduces a new method using image processing and YOLOv8 to detect crop diseases, improving accuracy and reducing the need for expert intervention.

## Contribution

A novel hybrid approach combining advanced image processing and YOLOv8 for efficient and accurate crop disease detection.

## Key findings

- The method achieved a recall of 0.94 and an overall accuracy of 92.567 in crop disease classification.
- Advanced image processing techniques improved image quality before disease detection.

## Abstract

The spread of plant diseases in important crops that influence the economy, particularly in Asia, such as tomatoes, coffee, cucumbers, olives, and wheat, poses a serious threat to agricultural production and global food security. Traditional detection methods are frequently labor‐intensive, slow, and lack the public availability of data, which subsequently impacts the model's generalizability and implementation in the real world for practical use. For this purpose, a computer‐aided approach is required to detect and classify diseases using crop images. In this research, images are initially processed using advanced image processing techniques like local contrast enhancement, wavelet transform, sigmoid correction, gamma correction, and median filtering, which are then evaluated using mean squared error and peak signal‐to‐noise ratio. After the processing phase, we utilize an advanced deep learning model, YOLOv8, to segment and classify crop diseases using publicly available data. This hybrid dataset includes data collection of 32 diseases. Using a large dataset, which comprises 32 diseases, to train our model, we implemented Transfer Learning using YOLOv8. We performed segmentation and classification with excellent recall and precision, with a recall of 0.94 and an overall accuracy of 92.567. The evaluation measures show dependable performance in crop disease identification across various circumstances. This will not only enhance the early disease detection in key crops but also reduce the intervention of experts, resulting in improved early disease diagnosis and the aversion of significant crop losses.

The performance of processed images is evaluated using mean‐squared‐error and peak‐signal‐to‐noise ratio. After the processing phase, an advanced deep learning model, YOLOv8, was used for the segmentation and classification of crop diseases. Using a large dataset comprising 32 diseases to train our model, we implemented Transfer Learning using YOLOv8. We performed segmentation and classification with excellent recall and precision, with a recall of 0.94 and an overall accuracy of 92.567.