Photo Annotation: A Study on Automatic Photo Management Based on Contextual Information

The increasing number of photos that people keep in their digital data storage devices often get scattered and it gets quite difficult to keep track of them effectively in time of need or to manage them efficiently. We present a process of photo annotation based on the metadata of the images and have developed a photo management application that derives information of location, daylight, weather, event name and persons related to the photo through external application programming interfaces (API) and libraries from the respective images metadata. These informations are annotated against the photos and saved into the database of the application so that the photos can be brought into the users view when they are sought according to these informations i.e. the location or event on which a particular photo was taken. We have also created a test database of 1350 images and our application has the accuracy of 82% for Daylight, 27.85% for events, 95% for locations, 40.75% for persons and 80% for weather data. The test images are taken from a span of two years and are of 40 events, 45 locations and of 25 camera models.