Download the best models based on accuracy from:
https://drive.google.com/drive/u/0/folders/1IPH_QMG21imbDQe9ieZ-qRumM_7aVThe
Download the Dataset from:
https://github.com/spMohanty/PlantVillage-Dataset
The challenge
Undeniably, the importance of the agriculture sector both for the global and local economies is enormous. That is the reason why, any challenges that appear and may influence the agriculture production need to be handled efficiently and timely. One of the biggest pain points that have been identified concerning the agriculture sector is the plant diseases that are evolving as the climate crisis is flourishing in our days. Not to mention that diseases are transferred globally more easily than ever before. Given that disease identification is mainly performed with visual methods, even today, from trained professionals and taking into consideration the spread of the diseases throughout the globe, it is impossible to handle all the separate circumstances arising at every corner of the earth.
The solution – Our project
Our project has the goal to overcome this obstacle by providing an easy tool with high accuracy that can be used throughout the world by anyone (from amateur gardeners to trained professional and farmers) and identify healthy from unhealthy plants and classify them based on their disease. By quickly and appropriately identifying the plant diseases throughout the world, substantial management issues and economic losses in the agricultural industry will be alleviated. Given that at least 10% of global food production is lost due to plant disease, the early identification and correct identification is of utmost importance.
Taking it a bit further
But identifying and classifying the disease of a plant is only the one side of the coin, or part of the solution. Our project aims at providing an end-to-end solution. Therefore, our tool aims at also providing which is the remedy for every separate disease identified. As a result, having only an image of a diseased plant, with the help of our project, the disease will be accurately identified, and the respective remedy will be proposed.
In a nutshell
This project is concerned with a new approach to the development of plant disease recognition model, based on leaf image classification using deep convolutional networks. The aim is to develop an easy but accurate system implementation in practice. All essential steps required for implementing this disease recognition model are fully described throughout the paper, starting from the dataset itself, up to the machine learning framework developed.
Our application
Our team has developed a User Interface (UI) to make the experience more interactive fast and efficient. With the designed UI, the user has just to upload the picture with the leaf whose disease needs to be identified and the rest will be made automatically. Specifically, the steps are listed below:
-
Upload the respective picture
-
The UI will output both the disease and the remedy proposal
-
In the upper left part of the screen, the user must upload the image with the diseased leaf. After the file has been browsed, at the right part of the screen the application outputs the following in the order that are listed:
- Your plant disease is: …
- Recommended Remedy: …
Finally, at the bottom left part of the screen the user can also populate the field with their location to have an overview about:
• Temperature
• Pressure
• Humidity etc.
This information may be of use, since a lot of diseases are linked with the weather conditions.
Short Term Goals
Given that most diseases generate manifestation in the visible spectrum, the naked eye examination of a trained professional is the primary technique adopted for plant disease detection even today. Undeniably, there are some diseases with no visible symptoms that need a more sophisticated analysis, but our team’s vision is to help and optimize the detection and decision-making progress in most of the diseases which are also visible to the naked eye. The goal of the specific project is twofold. The plant disease identification is a process that concerns two main steps that are consecutive. Only after successfully performing the two steps that are being listed below, the last step of overcoming the disease with the appropriate pesticides can take place:
1. Identify the exact plant that is being examined
2. Identify the disease of the plant
Given, that as a team we want to provide an end-to-end solution, the vision does not stop to the identification step. Having identified both the plant and the disease, the trained model will be able to also provide which are the most appropriate remedies (pesticides) to overcome the specific disease. In a nutshell, what we wish to accomplish is to provide an automated system designed to help identify plan from amateurs in the gardening process to trained professionals as a verification system in disease diagnostics. With a simple photo taken a three-way output will be produced:
• Plant identification
• Disease identification
• Appropriate remedies
Future Work
The main goal for the future work will be developing a complete system containing a trained model and an application for smart phone devices with features such as displaying recognized diseases in fruits, vegetables, and other plants, based on leaf images captured by the phone camera.
This application will serve as an aid to farmers regardless of the level of experience, enabling fast and efficient recognition of plant diseases and facilitating the decision-making process when it comes to the use of chemical pesticides.
Furthermore, future work will involve spreading the usage of the model by training it for plant disease recognition on wider land areas, where it would be possible to even identify the main cause of the plant disease such as worms, insufficient arable land properties etc.
By extending this research, we hope to achieve a valuable impact on sustainable development, affecting crop quality for future generations and optimizing the decision-making process given that both the disease identification and the route cause will be the combined output from a single solution.