Plantify
An AI-powered application that identifies plants from images using computer vision and machine learning.
- Type
- Computer Vision · AI Product · Product Design
- Role
- Product Designer / ML Engineer
- Contribution
- Product design · Model training · Evaluation
The problem
Most people cannot name the plants around them, and the usual routes to finding out, searching by description or browsing a field guide, are slow and depend on already knowing what to look for. A photograph is the most natural input: quick to take, and it holds the information an expert would use. Plantify set out to turn a single image into a confident identification, in an experience simple enough for someone with no botanical knowledge.
What I built
Plantify pairs a computer-vision model with a product experience designed around it. On the machine-learning side, an image classifier is trained to recognise plant species from photographs and evaluated on held-out images to measure how reliably it identifies them.
On the product side, the flow is built around the moment of capture: take or upload a photo, receive an identification, and see the result presented clearly enough to act on. Designing both halves together meant the interface could be shaped around what the model can and cannot do, rather than treating the model as a black box attached to a finished design.
Approach
From photograph to answer
Key decisions
Design around the model’s behaviour
The interface was shaped by how the classifier actually performs, rather than assuming every prediction is equally certain.
One photo, one clear answer
Capture is the whole interaction: the experience asks for a single image and returns a single, readable identification.
Measure on unseen images
Accuracy was measured on images held out from training, so the reported figure reflects identification rather than memorisation.
Outcome
- Plant-identification accuracy
- 90%+
During project evaluation, Plantify identified plants with over 90% accuracy, delivered through a capture-and-result flow designed for people with no botanical background.
- Capabilities
Working on something similar?
Start a conversation