Palm Fruit Vision

From photo to figures

From a single photo, Palm Fruit Vision estimates quality values that normally require an instrument or a laboratory. The model was built in-house on real bunches with reference measurements.

The photo

A sharp phone photo of the bunch. To build the dataset, each bunch was photographed from several angles.

The model

An in-house vision network analyses the image and estimates the four values. It rejects unusable photos (empty bunch, blur, off-topic).

The result

The four values and a harvest verdict, saved in your history with the photo.

Continuous improvement

Every new bunch measured in the field enriches the dataset; every new model version is evaluated on unseen bunches before it goes live.

What the model predicts

Oil content

Directly determines extraction yield.

Dry matter

Indicator of ripeness and water content.

FFA (free fatty acids)

Free acidity: freshness indicator, lower is better.

Protein

Nutritional value of the press cake.

Honest by design

A prediction from an image is not a laboratory analysis. It helps you decide, sort and prioritise at the right moment, not certify. We publish the accuracy we actually measure, and update it with every new model.