Summary
EgoGroups is a new benchmark dataset for social group detection—the task of identifying humans involved in reciprocal interpersonal interactions such as family members, friends, or customer-merchant pairs. Existing benchmarks suffer from low scene diversity and reliance on third-person sources like surveillance footage, limiting real-world evaluation of how groups form and evolve across cultural contexts. EgoGroups addresses this gap with a first-person (egocentric) dataset capturing social dynamics in cities across 65 countries, spanning low, medium, and high crowd-density settings and four weather/time-of-day conditions. The dataset includes dense human annotations of people and social groups, along with rich geographic and scene metadata. The authors use it to extensively evaluate state-of-the-art vision-language models, large language models, and supervised models on group detection. Developed by researchers including Jeffri Murrugarra-Llerena and Paola Cascante-Bonilla, the work is available on arXiv (2603.22249). It targets the social intelligence capabilities needed by agents operating in real-world environments.
Paper Overview
Field: Computer Vision
Authors: Jeffri Murrugarra-Llerena, Pranav Chitale, Zicheng Liu, Kai Ao, Yujin Ham, Guha Balakrishnan, Paola Cascante-Bonilla
arXiv: 2603.22249
Abstract
Social group detection, or the identification of humans involved in reciprocal interpersonal interactions (e.g., family members, friends, and customers and merchants), is a crucial component of social intelligence needed for agents transacting in the world. The few existing benchmarks for social group detection are limited by low scene diversity and reliance on third-person camera sources (e.g., surveillance footage). Consequently, these benchmarks generally lack real-world evaluation on how groups form and evolve in diverse cultural contexts and unconstrained settings.
To address this gap, the authors introduce EgoGroups, a first-person view dataset that captures social dynamics in cities around the world. EgoGroups spans 65 countries, covering low, medium, and high-crowd settings under four weather/time-of-day conditions. The dataset provides dense human annotations of people and social groups, as well as rich geographic and scene metadata.
Using this benchmark, the authors conduct an extensive evaluation of the social group detection capabilities of state-of-the-art vision-language models/large language models and supervised models.
---
*Auto-collected on 2026-03-25.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/177169020