Tesis

LabelMe: Online image annotation and applications

Citation

Torralba, A., B.C. Russell, and J. Yuen. “LabelMe: Online Image Annotation and Applications.” Proceedings Of the IEEE 98.8 (2010) : 1467-1484. Copyright © 2010, IEEE http://dx.doi.org/10.1109/JPROC.2010.2050290 Institute of Electrical and Electronics Engineers Final published version Fri May 25 20:39:50 EDT 2012http://hdl.handle.net/1721.1/61984 Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.

As Published Publisher Version Accessed Citable Link Terms of Use

Detailed Terms

INVITED PAPER

LabelMe: Online Image Annotation and Applications
By developing a publicly available tool that allows users touse the Internet to quickly and easily annotate images, the authors were able to collect many detailed image descriptions.
By Antonio Torralba, Bryan C. Russell, and Jenny Yuen

ABSTRACT | Central to the development of computer vision systems is the collection and use of annotated images spanning our visual world. Annotations may include information about the identity, spatial extent, andviewpoint of the objects present in a depicted scene. Such a database is useful for the training and evaluation of computer vision systems. Motivated by the availability of images on the Internet, we introduced a webbased annotation tool that allows online users to label objects and their spatial extent in images. To date, we have collected over 400 000 annotations that span a variety of different sceneand object classes. In this paper, we show the contents of the database, its growth over time, and statistics of its usage. In addition, we explore and survey applications of the database in the areas of computer vision and computer graphics. Particularly, we show how to extract the real-world 3-D coordinates of images in a variety of scenes using only the user-provided object annotations. Theoutput 3-D information is comparable to the quality produced by a laser range scanner. We also characterize the space of the images in the database by analyzing 1) statistics of the co-occurrence of large objects in the images and 2) the spatial layout of the labeled images. KEYWORDS | Image database; image statistics; object detection; object recognition; online annotation tool; video annotation;3-D Fig. 1. The entire data set from an early vision paper [36]. The original caption illustrates the technical difficulties of image digitization in the 1970s: ‘‘(a) and (b) were taken with a considerably modified Information International Incorporated Vidissector, and the rest were taken with a Telemation TMC-2100 vidicon camera attached to a Spatial Data Systems digitizer (Camera Eye 108)’’ [37].I. INTRODUCTION
In the early days of computer vision research, the first challenge a computer vision researcher would encounter would be the difficult task of digitizing a photograph [25]. Fig. 1 shows the complete image collection used for the study presented in [37] and illustrates the technical difficulties existing in the 1970s to capture digital images. Even once with a picture indigital form, storing a large number of pictures (say six) consumed most of the available computational resources. Today, having access to large data sets is fundamental for computer vision. Small data sets have the risk of overfitting by encouraging approaches that work only on a few selected cases. Moreover, it is hard to evaluate progress in the field with small data sets. This is specially relevantin
Vol. 98, No. 8, August 2010 | Proceedings of the IEEE 1467

Manuscript received April 16, 2009; revised April 1, 2010; accepted May 1, 2010. Date of publication June 10, 2010; date of current version July 21, 2010. This work was supported by the National Science Foundation Career Award (IIS 0747120) and by a National Defense Science and Engineering Graduate Fellowship. A. Torralba and J....