Crop Disease Detection from Phone Photographs
A phone application that identifies common diseases in a chosen crop from a leaf photograph and works without a network connection. The engineering challenge is that models trained on clean laboratory images collapse on real photographs taken in a field.
Intermediate
A small team, or one strong student willing to learn something new.1 semester, 2 students
5 to ship
3 optional extrasSuggested stack
What you should ship
- Trained classifier covering at least 8 disease classes plus healthy, for one crop
- Field validation set of photographs taken on phones in real conditions, separate from training data
- Quantised model running offline on a mid-range Android device with measured inference time
- Application with capture guidance, result, confidence and a low confidence path that declines to guess
- Comparison of accuracy on the clean dataset against accuracy on the field set
If you have time left
- Treatment guidance per diagnosis, sourced from an agricultural extension service and clearly attributed
- Offline history with sync when a connection is available
- Severity estimation as well as classification
The problem
Plant disease is often diagnosed late because the nearest expert is far away. A phone that gives a farmer a plausible diagnosis in the field is genuinely useful, and this problem is well suited to a student project because open datasets exist.
What you build
An image classifier for one crop, quantised to run on a phone offline, wrapped in an application that guides the user to take a usable photograph and reports a diagnosis with a confidence level.
The interesting result
The gap between laboratory accuracy and field accuracy. Public datasets in this area are mostly single leaves photographed against uniform backgrounds under even lighting. Real photographs have soil, shadows, other leaves, motion blur and inconsistent framing, and models routinely drop from ninety plus percent to something far less useful.
Measuring that gap, and then closing part of it with augmentation and your own field data, is a better project than another high number on a clean benchmark.
The hard part
Collecting a genuine field validation set. It requires going outside and taking photographs in the conditions the tool would actually be used in. Budget time for it and start early, since crop disease is seasonal.
Scope warning
One crop. Not five. And the low confidence path matters: a system that confidently misdiagnoses is worse than one that says it is unsure and recommends asking someone, particularly when the outcome is spraying a chemical.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Every session is free; a few slots open each week.
Follow along
New writing, resources and project ideas land here first.