Labeled dataset for integral evaluation of moving object detection algorithms: LASIESTA.
Computer Vision and Image Understanding(2016)
摘要
Complete database for assessing the quality of foreground detection strategies.Indoor and outdoor sequences with many categories addressing different challenges.All the sequences are fully annotated at both pixel and object levels.Information concerning stationary foreground objects.Sequences recorded with static and moving cameras. Display Omitted A public, complete, compact, and well structured database is proposed, which allows to test moving object detection strategies. The database is composed of many real indoor and outdoor sequences organized in different categories, each of one covering a specific challenge. In contrast to other databases, the proposed one is fully annotated at both pixel and object levels. Therefore, it is suitable for strategies exclusively focused on the detection of moving objects and also for those that integrate tracking algorithms in their detection approaches. Additionally, it contains sequences recorded with static and moving cameras and it also provides information about the moving objects remaining temporally static.To test its usefulness, the database has been used to assess the quality of some outstanding moving object detection methods.
更多查看译文
关键词
Database,Dataset,Ground-truth,Moving object detection,Foreground detection,Background subtraction,Challenges,Stationary foreground,Moving camera
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络