Few-Shot Object Detection: A Survey

Deep learning approaches have recently raised the bar in many fields, from Natural Language Processing to Computer Vision, by leveraging large amounts of data. However, they could fail when retrieved information is not enough fit vast number parameters, frequently resulting overfitting and therefore poor generalizability. Few-Shot Learning aims at designing models that can effectively operate a scarce data regime, yielding strategies only need few supervised examples be trained. These procedures are both practical theoretical importance, as crucial for real-life scenarios which either costly or even impossible retrieve. Moreover, bridge distance between current data-hungry human-like generalization capability. vision offers various tasks few-shot inherent, such person re-identification. This survey, best our knowledge first tackling this problem, focused on Object Detection, has received far less attention compared Classification due intrinsic challenge level. In regard, review presents an extensive description been tested literature, discussing their pros cons, classifying them according rigorous taxonomy.