On the Relationship between Head Pose, Social Attention and Personality Prediction for Unstructured and Dynamic Group Interactions
(Ramanathan Subramanian, Yan Yan, Jacopo Staiano, Oswald Lanz, Nicu Sebe)
Abstract
It is widely known that social attention and personality traits are dependent upon each other but the position of the head is more commonly used to determine the direction of attention in a small gathering/group. Unstructured and dynamic group was formed and captured via multiple cameras. Social attention features are excellent predictors of the Extraversion and Neuroticism personality traits. Results of the experiment showed that while prediction performance for both traits is affected by head pose estimation errors, the impact …show more content…
112 volunteers on Facebook were chosen who were then asked to complete two surveys online: the big five personality test (BFI-10) and the interpersonal circumplex (IPIP-IPC-32). Candidates with incomplete surveys were disqualified and about a 100 candidates were left. On this data, BoVW was implemented to extract features from images and the results were further refined using SIFT, SVD and VSEM. A t-test was run to compare traits to a set of about 100000 Facebook users randomly sampled from myPersonality. Six different classification algorithms were used: naïve bayes, support vectors, decision trees, logistic regression, nearest neighbors, rule learner. We ran a 60% training and 40% test split evaluation (f1-s) and provide f1-scores. We computed two baselines for the classification tasks: the majority baseline (maj) and the averaged majority baseline (avg maj). The majority baseline is close to 0.5, because we balanced the two classes for each task. The averaged majority baseline is the average of the performance obtained classifying all the instances before in one class and then in the other. Both the baselines were computed using …show more content…
This revealed that extrovert and stable people tend to have pictures where they are smiling and they appear with other people. Introverts tend to appear alone, neurotics tend to have images without humans and close-up faces. Not agreeable persons tend to use profile pictures with few colors. Profile pictures of people that is not open to experience tend to have strong spots of light and sometimes darkened figures. Affective people tend to smile in profile pictures. Dominant people tend to have very bright, visible or attractive pictures. Conscientiousness seems to be predictable by eye gazing
(Ramanathan Subramanian, Yan Yan, Jacopo Staiano, Oswald Lanz, Nicu Sebe)
Abstract
It is widely known that social attention and personality traits are dependent upon each other but the position of the head is more commonly used to determine the direction of attention in a small gathering/group. Unstructured and dynamic group was formed and captured via multiple cameras. Social attention features are excellent predictors of the Extraversion and Neuroticism personality traits. Results of the experiment showed that while prediction performance for both traits is affected by head pose estimation errors, the impact …show more content…
112 volunteers on Facebook were chosen who were then asked to complete two surveys online: the big five personality test (BFI-10) and the interpersonal circumplex (IPIP-IPC-32). Candidates with incomplete surveys were disqualified and about a 100 candidates were left. On this data, BoVW was implemented to extract features from images and the results were further refined using SIFT, SVD and VSEM. A t-test was run to compare traits to a set of about 100000 Facebook users randomly sampled from myPersonality. Six different classification algorithms were used: naïve bayes, support vectors, decision trees, logistic regression, nearest neighbors, rule learner. We ran a 60% training and 40% test split evaluation (f1-s) and provide f1-scores. We computed two baselines for the classification tasks: the majority baseline (maj) and the averaged majority baseline (avg maj). The majority baseline is close to 0.5, because we balanced the two classes for each task. The averaged majority baseline is the average of the performance obtained classifying all the instances before in one class and then in the other. Both the baselines were computed using …show more content…
This revealed that extrovert and stable people tend to have pictures where they are smiling and they appear with other people. Introverts tend to appear alone, neurotics tend to have images without humans and close-up faces. Not agreeable persons tend to use profile pictures with few colors. Profile pictures of people that is not open to experience tend to have strong spots of light and sometimes darkened figures. Affective people tend to smile in profile pictures. Dominant people tend to have very bright, visible or attractive pictures. Conscientiousness seems to be predictable by eye gazing