The Quividi precision reaches:
- 96% in detecting faces within the face detection distance. We hardly miss any face.
- 95% in attention time: we precisely track when someone's face is turned towards the video sensor with .1 second precision.
- 88% for gender attribution
- 75% for age attribution (in the four age classes) with our Pro version
Our Expert version provides an age that's within a +/-5 years range for 50% of the people analysed and within +/-10 years range for 95% of the people.
- 90% for mood estimate (Expert version)
- 80% for facial feature estimate such as glasses or beard (Expert version)
these scores are instant scores, ie out of 100 women faces being analysed, some 12 of them might be mistaken for men. But since about as many men might be mistaken for women, the final statistical result (or "pie chart") is very close to reality.
In about 5 to 7% of the cases, the solution might say that it is uncertain of the age and gender of the viewer (if one gender or age is unknown for a person, then both age and gender are reported as unknown).
For a 3rd party evaluation of the gender classification, contact your sales rep.