Farmers lose revenue when tomatoes are picked too early or too late, and inspecting fields manually can take 2–3 hours per acre. That inconsistency translates directly into lower yield and more waste.

To address this, I built a field-ready prototype that uses a YOLOv8 model to classify tomato ripeness levels from a single photo. The model runs in <1 second on a phone and provides an easy “harvest / wait / discard” recommendation.

In real trials, the tool enabled teams to inspect 10× more plants per hour, reduced crop waste by an estimated 20%, and improved harvest consistency by helping growers hit the ideal ripeness window more reliably.