LivePics-24: A Multi-person, Multi-camera, Multi-settings Live Photos Dataset

In 2024, the daily volume of photographs taken has soared to over 5 billion, with smartphones accounting for more than 94% of this staggering figure. At the forefront of this trend is the Apple iPhone, which has outshone traditional digital cameras to become the preeminent choice for photography enthusiasts worldwide. A standout feature of the iPhone is its Live Photos function, which not only captures a still image but also records a 1.5-second video snippet before and after the shutter is pressed, adding a vivid, dynamic layer to the photographic experience. Despite the widespread adoption and potential of Live Photos to elevate aesthetic quality and enrich user experience analysis, the research on Live Photos has been hampered by a dearth of publicly available datasets. To fill this void, our paper presents LivePics-24, an innovative dataset of live group images, compiled with informed consent from all depicted individuals. LivePics-24 is characterized by its variety, featuring an array of group compositions, genders, and ages, with participants positioned in multiple orientations and both masked and unmasked. This dataset is aimed to become a pivotal resource for researchers, facilitating the evaluation and refinement of computer vision algorithms tailored to live images and propelling advancements in areas such as face detection, emotional intelligence, and optimal key-frame extraction.