ChildLens: An egocentric video dataset for activity analysis in children

We present ChildLens, an egocentric video and audio dataset with detailed annotations for activities of naturalistic everyday experiences in children aged 3 to 5 years. A total of 109 h were recorded from 62 children in their home environment using a 140° wide-lens camera equipped with a microphone integrated in a child-friendly vest. Annotations include five location classes and 14 activity classes, covering audio-only, video-only, and multimodal activities. Good benchmark performance of two state-of-the-art models on the dataset—the Boundary-Matching Network for temporal activity localization and the Voice Type Classifier for detecting and classifying speech in audio—speak to the quality of the annotations. The ChildLens dataset will be freely available for research purposes via an institutional repository. It provides rich data to advance computer vision and audio analysis techniques and thereby removes a critical obstacle to studying the everyday context of child development, listed on the ChildLens website: https://www.eva.mpg.de/comparative-cultural-psychology/technical-development/childlens/ .