LeafShield

Sustain the Leaf, Empower the Farmer: Cultivating Sustainable Futures

Comment

Inspiration

Have you ever seen your plants’ leaves full of dark spots and have no idea what is happening to your plants? On a larger agricultural scale, farmers with much much larger crops are threatened with serious diseases that engender significant financial loss and in the long term, unsustainable agriculture.

Crop blight might not be omnipresent in highly developed countries, like here in the US, but in developing Asia countries, like ours, India and Vietnam, where agriculture employs more than 50% of the population, diseases are a big deal! Annually, crop losses worldwide range from 15-30%, in which Late Blight Disease is the most serious tomato disease with a worldwide loss of $6.7 Billion. This massive problem has inspired us to kickstart LeafShield, an application that helps gardeners and farmers detect the health condition of their plants based on their leaves and then provide feasible solutions to tackle the problem as soon as possible.

What it does

LeafShield is dedicated to gardeners and farmers who are busy with their work and have limited scientific knowledge about crop disease. They are, meanwhile, really in need of a quick solution to detect their crop’s state just by a SCAN and DETECT with a very high confidence level.

For the time being, LeafShield is available in both mobile app and web app, which is flexible and accessible to use. We now focus on analyzing tomato leaves and displaying up to 98% detection confidence level.

Challenges we ran into

  1. Creating a team
  2. Coming up with ideas
  3. Collection of data
  4. Building a model ~7 hours
  5. Technical challenges: exion issues, CRON issues, ML build issues, formation of data, formation of the request body for API
  6. Designing: limited knowledge of Figma, HTML, CSS

How we built it

We developed our application using Python, JavaScript, ReactNative for mobile apps, and ReactJs for web apps. Firstly, we trained a Machine Learning model using CNN (Convolutional Neural Network) based on the Plant Village dataset that includes nearly 2000 photos of leaves to detect the disease of a plant. Secondly, we developed API ( Application Programming Interface) which is built on the fastAPI framework where we are using post request. Then, we send requests as multipart/form-data and return JSON as a response. Later we created a web application using React JS where we used the drag and drop feature to upload an image. Then, we get the classification of disease as an output. Moreover, we created a mobile application for the same using react-native. To visualize the application and present its future potential, we use Figma to create a mockup.

Accomplishments that we're proud of

We've successfully built an ML model with an accuracy of 98.05%. We have built an API that can be used as an intermediate for our web application and mobile application. We built a web application on React JS and a mobile app on react-native using Expo Go and designed the mockup using Figma to realize our imagination. We've also successfully used Machine Learning to correctly detect the disease from just an image of an infected leaf. When we think about our uncles at home who can now detect the issue early, do the treatment promptly, and save their crops, we are proud that we can use our knowledge to contribute to their lives.

What we learned

Overall, we have better skills working with Figma, datasets, and backend development using ML, FastAPI, CNN, Python, JavaScript, ReactJs, and ReactNative. Besides, we are developing our teamwork skills, communication skills, and idea implementation skills. Not only does this project test our ability to develop an idea within a time limit, but it also motivates us to step out of our comfort zone and develop better technological applications in the future.

What's next for LeafShield

In the future, we hope to continue improving LeafShield with larger datasets, offering users more options for crops to scan and detect. We also plan to further develop current features to provide a better community for farmers, while making specialists more accessible for connection and support. Besides, as our dataset increases, there might be multiple potential diagnoses for the same scan. Hence, we want to develop our current confidence level feature to rank results in case of multiple diagnoses. We also want to add text language options for users’ accessibility, advice sessions from specialists, an AI Chatbot, and implement news API to offer users a variety of agricultural information.

In conclusion, through LeafShield, we hope to bridge the gap between agriculture and technology. Farmers can gain easy access to tech applications to directly identify what is going on with their plants, hence active solutions to crop problems and agricultural sustainability.

Built With

Share this project:

Updates