crop disease detector

• 

  showing the identified output

• 

  main page

Inspiration

Crop diseases destroy billions in agricultural output yearly. Early detection saves crops and helps farmers protect their livelihoods

What it does

Users upload leaf images. AI predicts disease (38 types) or healthy status with confidence percentage and treatment recommendation.

How we built it

Trained the ResNet50 model on the PlantVillage dataset (54K+ labeled images) using PyTorch. Built a web interface with Gradio for easy access.

Challenges we ran into

Model training timeouts in Colab. Solved by optimizing batch size and learning rate.

Accomplishments that we're proud of

• Trained a functional machine learning model in 8 days with 90%+ accuracy.
• Built and deployed a working Gradio web app demo.
• Learned PyTorch, transfer learning, and rapid deployment under deadline.
• Created a real solution that solves an actual agricultural problem.

What we learned

PyTorch CNN architecture, transfer learning, Gradio deployment, and rapid prototyping under deadline.

What's next for crop disease detection

Improve accuracy with longer training, add a treatment database, and add a mobile app version.

Built With

• python