Visual Sentiment and Recognition Tools

The objective for these tools is to detect objects and affectiveness in images and videos. It can be used for computer vision applications such as understanding the environment and will be the first time that computer has 'feeling' towards visual scenes. The result can be also used in other applications, such as social marketing or media & entertainment publisher where the marketers can predict viewers' sentiments and potentially customize for targeted marketing.

Challenges:

  • Object-based affective concepts in images need to be localized
  • Affective concepts are subjective, ambiguous and overlapping

Solution and results for publisher's sentiment:

Solution and results for viewer's sentiment:



This is a collaborative work of the Network Science Department in IBM Research and the Digital Video and Multimedia Lab in Columbia University.