MoodMe Face Analytics SDK gives you the insights on your audience’s engagement.
Clicks lie. Face don’t. That is, unless you play Poker.
Gender, Age & Ethnicity detection enable to target content to the right demographics.
The Face Analytics SDK provides the following data, for each image or video frame:
- gender: pink box for female, blue for male;
- age: age range & most probable age;
- expression: the strongest emotion (happy in the example) and a number (from 0 to 1) for each other basic emotion;
- attention: combines face orientation and engagement.
Applications: smart digital signage, where contents are personalized to segments of gender, age and groups.
Measure audiences Presence, Attention and Emotional engagement, wether in front of a large screen, smartphone or analyzing video footage after the experience has taken place.
Emotion recognition is the process of identifying human emotion from facial expressions automatically using computational methodologies.
MoodMe Emotions SDK is based on 7 different emotions: neutral, happy, surprised, sad, angry, fear, disgust.
Each of these are measured and ranked in a 0 – 100 range.
The combined data set for each emotion is a 7-dimension vector whose intensity gives the degree of each of these emotions.
The process of identification is divided into the first step, extraction of face characteristics of the individual. These characteristics are then matched with those of a data base, using a neural network. The results shown here are the matching of several Actors after loading our Face Data Base with all the faces of actors in IMBD Data Base (more than 20.000 actors).
Note the unmatched accuracy of the identification of actors whose heads are sideways, even turned at 90° from the camera.
Gender Age & Ethnicity Recognition
Gender Age & Ethnicity recognition from face images is an important application in the fields of retail advertising, marketing and security. MoodMe SDK is based on specialized filters for gender age & ethnicity recognition respectively which can be applied to 20 faces simultaneously.
MoodMe filters are constantly trained with broader & better quality data sets.
MoodMe SDK has demonstrated the effectiveness of the approach on its own datasets with 3960 training samples and achieved accuracies above 97%.
MoodMe Gender Age Ethnicity Emotions SDKs are being developed under a collaborative research project in deep learning and neural networks with 3 Universities: Rome La Sapienza, Rome Tor Vergata and Athens AIT.
Watch the team of Computer Vision researchers analyzed by their own algorithms.
These outperform approaches that rely on classifiers. Automatic gender age & sometimes ethnicity classification has become relevant to a broad range of applications. The use of neural networks delivers a significant increase in performance.