Explaining First Impressions: Modeling, Recognizing, and Explaining Apparent Personality from Videos

TL;DR

This paper explores explainability and interpretability in computer vision for human behavior analysis, focusing on modeling, recognizing, and explaining first impressions from videos, and introduces a related challenge and dataset.

Contribution

It is the first study to address explainability in first impressions analysis from videos and includes a new challenge, dataset, and evaluation protocol.

Findings

Analysis of explainability mechanisms in first impressions tasks

Introduction of a new dataset for explainability evaluation

Summarization of challenge results and future research directions

Abstract

Explainability and interpretability are two critical aspects of decision support systems. Within computer vision, they are critical in certain tasks related to human behavior analysis such as in health care applications. Despite their importance, it is only recently that researchers are starting to explore these aspects. This paper provides an introduction to explainability and interpretability in the context of computer vision with an emphasis on looking at people tasks. Specifically, we review and study those mechanisms in the context of first impressions analysis. To the best of our knowledge, this is the first effort in this direction. Additionally, we describe a challenge we organized on explainability in first impressions analysis from video. We analyze in detail the newly introduced data set, the evaluation protocol, and summarize the results of the challenge. Finally, derived from our study, we outline research opportunities that we foresee will be decisive in the near future for the development of the explainable computer vision field.