Return to knowledge base main page

How are the heatmaps generated by AI?

Our deep learning algorithm predicts the first 4-5 seconds of user attention on a new stimulus. It relies on pre-collected eye-tracking data to make these predictions. An algorithm can understand what attracts people’s attention in images by analyzing this eye tracking data on visual stimuli and analyzing those stimuli’s visual properties. 

Therefore, given a new image, a predictive eye tracking algorithm analyzes it and based on what it already learned from pre-collected data, it accurately predicts attention patterns on a new image. As a result, it can now mimic human vision with 90-96% accuracy compared to real eye-tracking studies but without data collection. 

The base of our technology is originally trained with ~70 000 image data sets from eye-tracking studies. We are also annually adding new data sets to ensure that our algorithm meets industry standards and stays the most accurate algorithm in the market. The image datasets we use are both open-source and proprietary. More about our technology you can find here.