UPlant

Inspiration

Not everyone knows how to take care of the plants and google search provides generic answers.\

What it does

UPlant lets you snap a photo of a plant leaf and get back the species name, a health assessment (healthy/unhealthy), and specific care recommendations based on detected symptoms.

How we built it

We trained a MobileNetV2 image classification model on Kaggle leaf datasets using TensorFlow/Keras. A FastAPI backend handles the pipeline: image in → Pl@ntNet API identifies the species → our model predicts health → rule-based logic returns care tips. The frontend is React Native + Expo so we can use the phone camera natively on iOS and Android.

Challenges we ran into

Not a lot of datasets on the unhealthy plants are available online. Deciding on the image classification model.

Accomplishments that we're proud of

We got a real end-to-end pipeline working: phone camera → species ID → health prediction → actionable recommendations and all that in one weekend.

What we learned

Leveraging an existing specialist API (Pl@ntNet) for species ID instead of training our own classifier and focus our ML effort on the health detection