Accuracy & Methodology

The app analyzes the visible object and scene in a photo, identifies likely object characteristics, and returns an approximate weight in your chosen unit. It infers from visual information; it does not sense mass directly.

Objects that look similar can have different materials, density, internal contents, or dimensions outside the frame. Those differences limit what any single image can reveal.

The confidence value is the model's indication that it could interpret the image and form an estimate. It is not a calibrated probability that the displayed weight is correct, an error margin, or a certification.

A high confidence signal should therefore never override the need for a real scale when the decision depends on exact weight.

Image quality affects the result. Retake the photo if the object is unclear or partly hidden.

• Show the whole object in sharp focus.
• Use even lighting and a simple, contrasting background.
• Avoid strong shadows, reflections, extreme angles, and heavy cropping.
• Keep unrelated objects out of the frame.
• Compare uncertain results with a physical scale.

Do not use a photo estimate for medical dosing, laboratory work, legal or trade measurement, regulated shipping, safety limits, valuable-material pricing, or precise recipes. Use an appropriate calibrated or certified scale for these cases.

We do not publish a universal accuracy percentage because performance varies by object category and capture conditions, and a single number could be misleading. If we publish a benchmark, it will identify the app/model version, reference scale, sample size, object categories, capture conditions, and the distribution of errors—not only an average.

Product claims and guidance on this site are reviewed by the Glipo product team. We label estimates as approximate, distinguish model confidence from measurement accuracy, and correct material errors when identified.