LeafLens

• 

  Diseased Leaf (upload to website from DemonstrationImages)

• 

  Healthy Leaf (upload to website from DemonstrationImages)

• 

  Diseased Leaf Website Results

• 

  Healthy Leaf Website Results

Inspiration

Looking outside at the variety of colors and shapes of leaves and plants, not knowing whether it's a dead plant or just a sign of fall coming.

What it does

LeafLens, your pocket botanist powered by AI. With just a quick photo of a leaf, LeafLens instantly detects whether the plant has a disease or is healthy, and provides a short explanation of what's going on.

How we built it

We built LeafLens using computer vision and generative AI. A TensorFlow-based CNN model (trained by us on the PlantVillage dataset) classifies plant leaf images as healthy or diseased. The predictions are then passed to OpenAI’s GPT-4o API, which generates simple explanations. The interface was built in Streamlit, allowing users to upload photos and instantly receive results in a clean, interactive dashboard.

Challenges we ran into

• A major challenge that we faced was trying to balance the accuracy of the predictions of the models using a limited dataset.
• Resizing leaf images while trying to maintain quality
• Trying to initiate TensorFlow properly . ## Accomplishments that we're proud of
• We made and trained an AI model without cloud computing
• Having 96% accuracy when training data

What we learned

• How to train and test an AI model
• How to build a website
• How to implement OpenAI

What's next for LeafLens

• Expanding the model to make it more accurate and reliable when interpreting pictures
• Make a mobile app where the user can capture an image of a leaf within the app and receive information.

Built With

• openai
• python
• streamlit
• tensorflow