Machine learning · Agriculture
LeafLens.
A crop-health prototype that turns leaf images into paddy disease and tea-quality assessments.
- Status
- Demonstrated prototype
- Context
- Agricultural image classification
- Technology
- Python · TensorFlow · Keras · Streamlit · Docker

Problem
Farmers need understandable information about visible crop damage and leaf quality. A model alone is not a usable interface for that task.
Solution
LeafLens combines image classification with a Streamlit application for paddy disease and tea-quality assessment. The project includes model-training notebooks, model assets, and packaging for local use.
Architecture
- 01Leaf image
- 02Image preprocessing
- 03TensorFlow / Keras model
- 04Streamlit assessment
My contribution
- Worked on the transfer-learning image classification pipeline and the Streamlit assessment interface.
- Packaged the prototype for local use with Docker and a standalone executable.
Engineering challenge
The engineering work extends beyond training: image preparation, model loading, and presenting the result all need to work together in an application people can use.

Results & demo
One Streamlit app demonstrates two workflows: paddy disease classification and tea quality assessment.
Short LeafLens interface preview.
Watch demonstration (opens in a new tab)What I’d improve
Validate both classifiers on independent field images and report per-class results.
Presented as an agricultural AI prototype. No independently verified field accuracy, yield improvement, or deployment impact is claimed. Image-based results need agricultural validation.