Sign Spotting

Sign spotting is the task of automatically locating instances of specific signs within a continuous signing video, as opposed to classifying a pre-segmented isolated sign. It is a building block for search-by-sign in archive footage, automatic captioning of signed media, and lexical-resource construction, and is technically harder than isolated sign language recognition because it must handle coarticulation between signs, depicting signs, and the absence of clear word boundaries. In accessibility research, robust sign spotting matters because it is the gateway to making large amounts of existing signed video content searchable and indexable for Deaf users in their first language.