INTRODUCTIONUnconsciously, humans evaluate situations based on environment and social parameters when recognizing emotions in social interactions. Contextual information such as the ongoing task, the identity and natural expressiveness of the individual, and other people involved, helps us interpret and respond to social interactions [1]. Without context, even humans may misunderstand the observed facial, vocal or body behavior. Then, an important related issue that should be addressed in automatic affect recognition is how to take into account the context information for real-world affect related applications.Investigating context in affect recognition is relevant for the fields of Affective Computing and Intelligent Interaction since contextual information cannot be discounted in doing automatic analysis of human affective behavior [1]. Embedding contextual information, such as culture and environment, provides a different flavor to each interaction, and makes for an interesting scientific study. Such kinds of analysis lead us to consider real-world parameters (e.g. medical environment) and complexities in developing humancentric systems for affect recognition.The aim of this Workshop is to explore new challenges in automatic context based audio, visual, body and/or multimodal affect recognition. The key aim of the workshop is to explore the issues, benefits, and drawbacks of integrating context on affect production, interpretation and recognition. We wish to investigate the cutting-edge behavioral studies and methodologies that can be applied to (1) model the social and cognitive theories of context based social and/or affective interaction (2) automatic extraction of context information (like multi-modal sensing systems, observation, and behavior models), (3) incorporation of contextual information in emotion corpora (e.g. how it ought to be represented, what contextual information are relevant (i.e. is it domain specific or not)), and (4) integration of the context to the audiovisual frameworks for affect recognition to improve their performances.The workshop focuses on making affect recognition more robust and deployable in real-world situations (e.g. work, home, school, and health care environment) by focusing on simple domains of applications and analyzing how affect recognition systems can be improved via contextual information.As affective computing is a multidisciplinary field that requires theories, methods, and technologies from different disciplines, the Workshop aims at bringing together scientists working in related areas of social signal processing, psychology, affective computing, ambient computing and smart environments, and machine learning to share their expertise and achievements in the emerging field of automatic and context based audio, visual and/or multimodal affect analysis and recognition.The contributions of the second edition of the context based affect recognition workshop tackle contextual information in relation to corpus recording and annotation, the role of context in affect rec...