Predicting Group Cohesiveness in Images

Abstract

The cohesiveness of a group is an essential indicator of the emotional state, structure and success of a group of people. We study the factors that influence the perception of grouplevel cohesion and propose methods for estimating the humanperceived cohesion on the group cohesiveness scale. In order to identify the visual cues (attributes) for cohesion, we conducted a user survey. Image analysis is performed at a group-level via a multi-task convolutional neural network. For analyzing the contribution of facial expressions of the group members for predicting the Group Cohesion Score (GCS), a capsule network is explored. We add GCS to the Group Affect database and propose the 'GAF-Cohesion database'. The proposed model performs well on the database and is able to achieve near human-level performance in predicting a group's cohesion score. It is interesting to note that group cohesion as an attribute, when jointly trained for group-level emotion prediction, helps in increasing the performance for the later task. This suggests that group-level emotion and cohesion are correlated.

D2C-Based Hybrid Network for Predicting Group Cohesion Scores

“…Furthermore, Treadwell et al [9] created GCS (shown in Fig. 1) to evaluate cohesiveness among group members based on a 4-point scale of strongly disagree (SD), disagree (D), agree (A) and strongly agree (SA), or can be represented by numbers from 0 to 3, correspondingly [10]. These studies have formed the fundamental knowledge for the evolution of social psychology especially, group cohesiveness.…”

“…This is one of the first studies that utilizes machine learning to predict group cohesion of a group of people in videos. Recently, the EmotiW 2019 presented the first study of group cohesion prediction in statics images [10,11]. The EmotiW organizers provided a database for group cohesion, which is an upgrade version of Group Affect Database (GAF 3.0) [20] with group cohesion labels (GAF-Cohesion).…”

“…The available resources motivate researchers to build an automatically estimate group cohesion system. Ghosh et al [10] provided a database for group cohesion that was applied in the Group Cohesion sub-task of the 7 th Emotion Recognition in the Wild (EmotiW 2019) Challenge [11]. In Ghosh's paper, two approaches were proposed to build models for GCS prediction, namely image-level (scenelevel) and face-level.…”

“…We developed the Discrete labels to Continuous scores (D2C) method to predict labels in continuous dimensional from discrete categorical. The method's idea is an improvement of the joint training method in Ghosh's paper [10] where discrete labels directly participate in generating continuous labels instead. In this case, discrete labels and continuous labels were group emotion and group cohesion labels, correspondingly.…”

GAME-ON: A Multimodal Dataset for Cohesion and Group Analysis

“…Ghosh et al proposed methods to automatically predict group cohesiveness in images from the GAF 3.0 dataset, focusing on facial expressions [54]. Their approach achieves near human-level performance in predicting a group's cohesion score, but this mainly concerns perceived cohesion and does not provide any insight on the underlying behavioral dynamics of cohesion.…”