Crop Yield AI

Help

How to check a field, read the result and keep your history.

Getting started

  1. Add photos. On Analyze, drop up to 15 top-down photos of your fields, or add a sample. Each one gets a plot number, P-01 onwards.
  2. Check yield. Press Check yield. Each photo is shrunk to 1200 px in your browser, then the server measures it and predicts the yield in tonnes per hectare.
  3. Compare plots. Your plots line up from highest to lowest yield. Select one to see its rating, recommended crops and the colour map of its plants.
  4. Add the location. On a result, choose Add location and enter the field's latitude and longitude to include today's weather and the soil type.
  5. Keep your history. Every result is saved in History. Sort it, open any record again or export it as a CSV file.

Questions

Which photos work best?

Straight-down photos of the crop during the growing season, from a drone or a satellite map. Angled photos, sky and close-ups of single leaves give unreliable results.

How is the yield worked out?

Every pixel is scored for greenness with a colour index that needs only a normal RGB photo. Five measurements from those scores go to a trained model, which predicts tonnes per hectare. Extra checks stop water, sand and bare ground from being rated as productive crops.

What do Good, Fair and Poor mean?

Good is 7 tonnes per hectare or more, Fair is 3 to 7, and Poor is under 3.

How are crops recommended?

A model trained on Malaysian crop data ranks ten crops: paddy, oil palm, rubber, durian, pineapple, coffee, papaya, cotton, banana and corn. With a location it also weighs temperature, rainfall and soil pH.

What is the colour map on a result?

Switch a result from Photo to Colour map to see each pixel coloured by its greenness score, from brown for bare soil to deep green for dense canopy. It is drawn in your browser with the same formula the server uses.

Where is my data stored?

Your history stays in this browser. If you sign in with Google, results are also saved to the project's database under your email so they follow you to other devices. Uploaded photos are kept on the server so a result can be re-run with a location.

Which coordinates should I enter?

Decimal degrees, for example 3.1390 and 101.6869 for Kuala Lumpur. In Google Maps, right-click the field and click the numbers to copy them.

Can I change the units?

Yes. In Settings, choose tonnes per hectare, kilograms per hectare or bushels per acre.

About Crop Yield AI

What it is
A Final Year Project by CHENG QING HONG that predicts crop yield from ordinary field photos.
Weather
Today's forecast from the Open-Meteo service for the field's location.
Soil
Soil texture and pH from SoilGrids by ISRIC World Soil Information.
Version
v2.0. Built with Python Flask, scikit-learn and OpenCV.