All projects

Fruit & Veg Detector — Live

point your camera at a fruit, watch the model name it — and see where it looked

TensorFlow.jsMobileNetV2Transfer Learningruns in-browser201 classes

This runs the actual trained image model right in your browser — no server, and no photo ever leaves your device. A MobileNetV2 network (transfer-learned on the Fruits-360 dataset, 201 classes) looks at a photo or webcam frame and ranks its guesses. Toggle the heat map to see exactly which pixels drove the decision.

Loading the model (~4 MB, one-time)…

Read the write-up: how this model works and where it breaks →