Efficient visual search in images and videos

Visual search has been a fundamental research topic in computer vision. Given a query, visual search aims to find the query’s relevant items from the database. Accuracy and efficiency are two key aspects for a retrieval system. This thesis focuses on the latter one, the search efficiency. To be specific, we address two basic problems in efficient visual search: 1) deep compact feature learning and 2) fast object instance search in images and videos.Doctor of Philosoph