MoodMe owns proprietary data bases and uses public data sets to apply semi – supervised data quality improvement. These data sets are annotated with those face characteristics of interest: landmarks, emotions (7 base emotions plus subtle emotions and micro expressions as of 2020), gender, age, ethnicity as well as recognition/identification and detection of face attributes (beard, mustache, hat, eyewear).

We optimize Neural Networks and applies the most recent compression techniques such as pruning. MoodMe combines the best proprietary neural networks with open frameworks like TensorFlow.