Recognizing Families In the Wild (RFIW)
It is with pleasure we announce the fifth and final large-scale kinship recognition data competition, Recognizing Families In the Wild (RFIW), in conjunction with the 2021 FG. RFIW has been made possible by releasing the largest and most comprehensive image database for automatic kinship recognition, Families in the Wild (FIW).
All submissions (i.e., challenge papers, general paper submissions, and Brave New Ideas) will be peer-reviewed for publication as part of RFIW2020 in the IEEE International Conference on Automatic Face and Gesture Recognition (AMFG) proceedings. Also, we expect authors will join the workshop during the 2021 AMFG conference on 15–18 December (Virtual Event, Jodhpur, India).
For more details click here.
Contextual Emotion Recognition Challenge
As society is moving rapidly towards more and more technological advancements with an exponential speed, there has been a conscious effort to learn the more ‘human factors’ involved in these innovations. Multiple works have been emerging on these lines, combining the fields of human cognition and technology together. Emotions are intrinsic to every human being and whether consciously or subconsciously, there is one or the other emotion behind every decision made by humans. In order to assess those processes involving emotions, any technological product being built needs to be backed by a thorough understanding of the same in order to be interactive and suit the needs of humans. Hence, there is a need to build systems that are able to recognise emotions from humans holistically, of which context is an inevitable aspect.
Emotion, affect, and behaviour recognition via computer vision has been deeply associated with facial expressions, and the detection of emotions has primarily been based on facial images. However, spatiotemporal context, including action, interaction with others, and place, plays an imperative role in emotion recognition. However, such contextual information cannot be incorporated using facial images. Further, research on visual sentiment analysis methods is also not feasible for identifying salient context information, mainly due to the efforts involved in encoding such contexts. In addition, while differential processing and attentional styles have been identified between western and eastern cultures , emotion recognition data sets containing any such contextual information are entirely absent for the Indian population. Therefore, we contribute to this research gap by introducing Indian Contextual Emotion Recognition (ICER) data set inclusive of video clips only pertaining to the Indian context.
This challenge contains two sets of videos downloaded from official YouTube channels with creative commons license. Both the data sets contain 2-10 second long video clips with a single subject in focus and may or may not include multiple camera angles of the same. The first set is the strongly annotated data set consisting of 4,000 video clips, each rated on a Likert scale of 5 points for the 7 Emotion Classes namely Happiness, Sadness, Anger, Fear, Disgust, Surprise, and Neutral for the categorical model of emotions. Similarly, the Valence and Arousal Ratings for the dimensional model have also been annotated using a Likert scale of 5 points. On the other hand, the second set is the data set of 7,000 video clips with no labels. The goal of this challenge is two-fold. The first part of the task is to perform a classification task on the strongly annotated data set. The second part of the task is to perform annotation for the second set of data, using the best-performing model from the first part of the task. Currently, there is no baseline. This challenge is open to all participant demographics.
For more details click here.
We solicit proposals for conducting competitions/hackathons for the 2021 IEEE Conference on Automatic Face and Gesture Recognition (FG 2021).
The goal of competitions is to mobilize the research community to solve a challenging problem of relevance to FG. For example, the participants can be expected to compete on a machine learning task of interest to the FG community where the problem and datasets are defined and released by the organizers of the competition.
The goal of hackathons is to encourage fast-paced innovation in an inclusive and agile manner and encourage development of working prototypes/demos that solve a pressing technology problem or a use case of relevance to the FG community.
The proposals may be targeted toward a specific participant demographic (e.g., undergraduate, or graduate students). Interdisciplinary topics that will attract a significant cross-section of the community are especially encouraged.
Competitions/Hackathon track will be held alongside the main conference (December 15 to December 18, 2021), where the results will be discussed/winners announced.
Competition/Hackathon proposals must be sent in PDF format (max. 3 pages) to the Competition/Hackathon Co-chairs, Abhinav Dhall (abhinav.dhall@monash.edu) and Sunpreet Singh Arora (sunarora@visa.com) with email subject “FG 2021 Competition/Hackathon Proposal”. The proposal must include the following information: