Plant disease diagnosis from images

The problem

A farmer sees a yellowing leaf without knowing why: thirst, a fungal infection, or iron deficiency? Each has a completely different treatment, and the wrong one wastes time and makes things worse. Harder still, after treatment the farmer can't tell whether the plant actually improved, since visual memory is limited and change is slow. Following up with photos turns a subjective impression into a real comparison.

How it works

1. ImagingThe user uploads a photo of the leaf or plant through the website.
2. AnalysisAn AI model classifies the condition and gives a confidence level.
3. ExplanationThe system explains why it suspects the condition and the suggested action.
4. ContextHumidity, light, and temperature sensors add readings that support the diagnosis.
5. Plant recordEach plant gets a record that builds up its photo history.
6. ComparisonNew photos are compared with old ones to measure improvement or decline.

Project features

• Follow-up over time is the system's core value — it doesn't stop at classifying a single photo.
• It combines computer vision, sensors, and a web interface.
• It can be demonstrated live: photograph a leaf in front of the committee and show the result immediately.
• It belongs to Jordan's agricultural sector and has real users.
• Linking the diagnosis with sensor readings improves the result's accuracy.

Expected challenges

These are the main obstacles to expect during the build, laid out clearly so you know them before you start.

• Ready models exist for plant diseases, but they're trained on crops not all of which grow locally.
• Lighting affects the leaf's color in the photo, so the diagnosis depends on when the photo is taken.
• Nutrient deficiencies and fungal diseases are hard to tell apart because their symptoms look alike.
• Comparing two photos of the same plant requires a consistent angle and lighting.
• The confidence level must be shown transparently: a confident wrong diagnosis is worse than admitting uncertainty.

The consultation gives you a solution for each one.

Expected committee questions

Prepare your answers in advance: these questions come up often in defenses of similar projects.

1. How many images were used for training, and how were diseases with few images handled?
2. How accurate is the model on field photos compared with lab photos?
3. How does the system measure improvement between photos of the same plant?
4. What happens if the user photographs a plant or disease the model doesn't know?

Practice the defense, question by question