Most wrong answers from a photo are decided before the photo is taken. A tilted phone, a heap or bad light makes a good sample look bad, or a bad one look good.
Image-based grading counts and measures every visible grain, so the photo must show every grain clearly, at a known scale, from directly above.
| Mistake | What it does to the result |
|---|---|
| Phone tilted | Sizes wrong across the photo |
| Grains heaped | Count too low; hidden defects missed |
| No scale | No millimetre sizes or weight basis |
| Shiny tray | Reflections confused with grain |
| Too few grains | Percentages swing with each grain |
If the photo cannot give a trustworthy answer, the right response is to ask for another one, with the reason, not to produce a number anyway. A refused photo costs seconds; a wrong grade costs a dispute.
GrainGrade turns one phone photo of a wheat sample into a grain count, sizes in millimetres, defect percentages and the FCI FAQ grade, with a range on every number instead of a bare decimal. The camera coaches the shot first (level, a scale marker in frame, focus and light) and refuses a bad photo with a reason in Hindi or English rather than analysing it into a wrong answer. Seven defect classes are checked, grading rules are data so a mill’s or exporter’s own spec can sit beside FCI FAQ, and every sample is saved to its lot with a shareable PDF or WhatsApp report. GrainGrade is in development: counting is built and measured; the other figures on the product page are design targets. Moisture is not visible to a camera and is entered from a meter reading.
From directly above with the phone level, grains spread in one layer on a plain matte background, a scale marker in frame, sharp focus and even light.
An angled photo makes nearer grains look larger than farther ones, so sizes and some defect calls go wrong.
It gives the scale, so grain sizes can be reported in millimetres.
See it on your own data. GrainGrade — Wheat graded from a photo. Book a 30-minute working session with an engineer.
