Description
The term “emotion recognition” is rather broad and can cover anything from identifying the emotion of a single face to inferring the emotions of an entire group. The term “sentiment analysis” is also broad and can refer to any number of different techniques for evaluating how positive or negative a statement might be.
Challenge
The client is in the customer relationship business. They needed to know if the customer’s expressions changed from the time they entered the store and by the time they left the store. They needed an AI system to capture one or more frames for each customer as they go through the store, and again as they leave. They also needed a way to train their algorithm to study these expressions in order to improve their customer relationships. They already had CCTV and DVR reporting installed in the store.
Solution
The company called us and asked us to recommend an AI/ML solution for their problem. We recommended that we would use sentiment analysis for this project because it is capable of extracting sentiments from facial recognition systems. We trained a database of Caucasian people in various expressions. This database was also followed by other race and cultural faces in all possible expressions. Using these models generated, which scanned the video feeds (live/ offline) for face detection and storing the relevant data on the cloud.