查看文章

Despite the increasing number of successful related work- shops and panels, lifelogging has rarely been the subject of a rigorous comparative benchmarking exercise. Following the success of the new lifelog evaluation task at NTCIR-12, the first ImageCLEF 2017 LifeLog task aims to bring the attention of lifelogging to a wide audience and to promote research into some of the key challenges of the coming years. The ImageCLEF 2017 LifeLog task aims to be a comparative evaluation framework for information access and retrieval systems operating over personal lifelog data. Two subtasks were available to participants; all tasks use a single mixed modality data source from three lifeloggers for a period of about one month each. The data contains a large collection of wearable camera images, an XML description of the semantic locations, as well as the physical activities of the lifeloggers. Additional visual concept information was also provided by exploiting the Caffe CNN-based visual concept detector. For the two sub-tasks, 51 topics were chosen based on the real interests of the lifeloggers. In this first year three groups participated in the task, submitting 19 runs across all subtasks, and all participants also provided working notes papers. In general, the groups performance is very good across the tasks, and there are interesting insights into these very relevant challenges.