A CNN classifies plant disease from a leaf image in seconds, with three disease-specific prevention tips per prediction.
Source available · Not hosted
PythonTensorFlow/KerasOpenCVDjangoHTML/CSS
// Problem
Diagnosing crop disease from a leaf means manual visual inspection of symptoms — slow and easy to get wrong. Upload a photo instead and get a classified disease with confidence in seconds.
// How I solved it
Built a Django web application around a CNN classifier: a user uploads a leaf photo and receives the predicted disease plus confidence in under two seconds — replacing manual visual inspection of symptoms.
Trained and evaluated the CNN on a labelled crop-image dataset; preprocessing (resize, normalisation, augmentation) to hold accuracy on unseen leaves.
Paired every prediction with three disease-specific prevention tips, so the output is actionable advice rather than a bare label.
// Features
Django web app around the CNN — upload a leaf photo, get disease class plus confidence.
Preprocessing pipeline — resize, normalisation, augmentation — to hold accuracy on unseen leaves.
Three disease-specific prevention tips attached to every prediction.
Browser upload form as the single input; result panel shows class, confidence and tips.
// Result
Leaf photo to predicted disease plus confidence in under two seconds — replacing manual visual inspection — with three disease-specific prevention tips attached to every prediction.
Notes: CNN built with TensorFlow/Keras; OpenCV used for image handling. Not hosted as a live product — clone the repository for source and setup instructions.
// Media
01 · Home — three steps from leaf photo to treatment recommendation.
02 · Detection entry — Scan Now opens the upload flow.03 · Result — Tomato Late Blight at 88% confidence, three prevention tips.
Screen recording · detection flow14.1 MB · loads on demand
