No room for error

To measure accurately, good tools are essential. And for those tools to function correctly, they need to be properly calibrated.

That’s why each year, usually during the summer, we calibrate our tools. This allows us to understand their margin of error and consider the precision with which we obtain our data.

A few months ago, we explained how we calibrate our tools in this blog post here.

Our most recent calibration took place in August 2024, yielding very positive results. The margins of error were minimal or nearly nonexistent in key indicators for our projects, such as eye tracking, and the detection of behaviors and emotions.

The maximum error is higher in areas like age estimation or gender differentiation, but who hasn’t made a mistake guessing someone’s age?

Here are the calibration results from August 2024:
• Maximum eye tracking error: 0.2 mm
• Maximum counting error: 0.4%
• Maximum age estimation error: 10.1%
• Maximum gender estimation error: 7.9%
• Maximum emotional identification error: 0.6%
• Maximum flow direction error: 0.01%
• Maximum behavior detection error: 0.9%