krishi

A website that uses a neural net to analyze images of plants and diagnose common diseases. It also provides an example image and info about each condition including symptoms, treatment, and much more.

Comment 8

Inspiration

We were inspired after watching videos about the monoculture of Cavendish bananas and how Panama disease is wiping out large swaps of the crop across Asia and South America. Farmers are being forced to just burn down their plants in hopes of limiting spread. We hope that this is a useful tool that benefits farmers. The name krishi comes from the Sanskrit word for agriculture.

What it does

The user uploads images of the fruit or leaf of one of the compatible plants, and our neural net identifies what type of disease (if any) the plant has. It then displays all relevant info for that disease. If the image uploaded is not a compatible plant leaf or fruit, our app will instead show an epic Bernie x Psy meme.

How we built it

We used Flask and Python to allow users to upload images, and we trained our neural net with the Identification of Plant Leaf Diseases dataset from Mendeley Data and Tensorflow. We then used HTML, Jinja, and CSS to build out our display pages. Our model is too large to upload to Github, but you can download a pre trained model into the Final/static folder from the Google Drive link here: https://drive.google.com/file/d/1cv09M407sOdQDwhL5ryXoMjRMYlGkMu8/view?usp=sharing. This model is trained on 15 epochs and must be named '15epoch.h5' for the code to work.

Challenges I ran into

Tensorflow requires Python 3.8 or earlier, and we couldn't get it to install with a virtual environment. Luckily, PyCharm came to the rescue by making the entire run process significantly easier than in VSCode (we still used VSCode to edit because PyCharm is so ugly lol).

Accomplishments that we're proud of

  • Using a Bernie meme
  • Building a functioning neural net
  • Compiling a lot of useful plant information
  • Getting through some ugly plant pics
  • Designing a 7/10 looking website (if Blake Shelton can win sexiest man alive, our website can win prettiest)

What we learned

  • How to upload and save images in Flask
  • How to improve our neural net through additional training epochs and the ML equivalent of whey

What's next for krishi

  • Helping us figure out whether the plants from our own gardens are safe to eat
  • Training our neural net even harder like in Rocky 2 (we even named her Krishi Balboa)
  • Hosting our site with a custom domain

Built With

Share this project:

Updates