Mining from large image sets

摘要: So far, most image mining was based on interactive querying. Although such querying will remain important in the future, several applications need at wide scales that it has to run automatically. This adds an additional level problem, namely apply appropriate further processing different types of images, and decide automatically as well. paper touches those issues we discuss landmark images coming from webcams. The first part deals with automated collection landmarks, which are then also annotated enriched Wikipedia information. target application is users photograph landmarks their mobile phones or PDAs, get information about them. Similarly, can photo albums object interest be delineated images. pipeline propose actually retrieves more than manual keyword input would produce. second entirely source data, but one produces massive amounts (although typically not archived): They produce a single location, rather continuously over extended periods time. We approach summarize data handling quite applied