VAISL: Visual-Aware Identification of Semantic Locations in Lifelog