The PASCAL Visual Object Classes (VOC) challenge：（帕斯卡的视觉对象类(VOC)的挑战）.pdfVIP

Int J Comput Vis (2010) 88: 303–338 DOI 10.1007/s11263-009-0275-4 The PASCAL Visual Object Classes (VOC) Challenge Mark Everingham · Luc Van Gool · Christopher K. I. Williams · John Winn · Andrew Zisserman Received: 30 July 2008 / Accepted: 16 July 2009 / Published online: 9 September 2009 © Springer Science+Business Media, LLC 2009 Abstract The PASCAL Visual Object Classes (VOC) chal- 1 Introduction lenge is a benchmark in visual object category recognition and detection, providing the vision and machine learning The PASCAL 1 Visual Object Classes (VOC) Challenge con- communities with a standard dataset of images and anno- sists of two components: (i) a publicly available dataset tation, and standard evaluation procedures. Organised annu- of images and annotation, together with standardised eval- ally from 2005 to present, the challenge and its associated uation software; and (ii) an annual competition and work- dataset has become accepted as the benchmark for object shop. The VOC2007 dataset consists of annotated con- detection. sumer photographs collected from the flickr2 photo-sharing This paper describes the dataset and evaluation proce- web-site. A new dataset with ground truth annotation has dure. We review the state-of-the-art in evaluated methods for been released each year since 2006. There are two princi- both classification and detection, analyse whether the meth- pal challenges: classification—“does the image contain any ods are statistically different, what they are learning from the instances of a particular object class?” (where the object images (e.g. the object or its context), and what the meth- classes include cars, people, dogs, etc.), and