The Oxford Buildings Dataset图像检索数据集

The Oxford Buildings Dataset

James Philbin, Relja Arandjelović and Andrew Zisserman 

Overview

The Oxford Buildings Dataset consists of 5062 images collected from  Flickr by searching for particular Oxford landmarks. The collection has been manually annotated to generate a comprehensive ground truth for 11 different landmarks, each represented by 5 possible queries. This gives a set of 55 queries over which an object retrieval system can be evaluated. 

Groundtruth Queries

The following image shows all 55 queries used to evaluate performance over the ground truth. 

Flickr queries used to download images All Souls Oxford Balliol Oxford Christ Church Oxford Hertford Oxford Jesus Oxford Keble Oxford Magdalen Oxford New Oxford Oriel Oxford Trinity Oxford Radcliffe Camera Oxford Cornmarket Oxford Bodleian Oxford Pitt Rivers Oxford Ashmolean Oxford Worcester Oxford Oxford

For each image and landmark in our dataset, one of four possible labels was generated:

1. Good - A nice, clear picture of the object/building.
2. OK - More than 25% of the object is clearly visible.
3. Bad - The object is not present.
4. Junk - Less than 25% of the object is visible, or there are very high levels of occlusion or distortion.

Database Rights

The Oxford buildings dataset consists of images provided by "Flickr". Use of these images must respect the corresponing terms of use. The identity of the images in the database has been obscured. Any queries about the use or ownership of the data should be addressed here.

Downloads

All data needed for evaluation is given below:

1. 5k Dataset images
2. Groundtruth files
3. C++ code to compute the ground truth
4. Additional 100k distractor images (Flickr 100k)

Additionally, we've made extra data available for the 5K dataset:

1. README for the following files
2. Compressed binary file of SIFT descriptors for the 5K dataset [MD5: fbc0e85c5065f6d97d519a7f2ed3e3f9]
3. Compressed text files containing word IDs and geometry for the 5K dataset using a vocabulary size of 1M
   [MD5: 0734e6023f6bf2140b9531af22ce7953]

Computing the average precision

1. Compile the compute_ap.cpp file, using (on linux) g++ -O compute_ap.cpp -o compute_ap.
2. To compute the average precision for a ranked list, rank_list.txt, one runs ./compute_ap christ_church_1 rank_list.txt, for the first Christ Church query, etc.

Relevant Publication

Philbin, J. , Chum, O. , Isard, M. , Sivic, J. and Zisserman, A.
Object retrieval with large vocabularies and fast spatial matching
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2007) 
Bibtex source | Abstract | Document: ps.gz PDF

Acknowledgements

This work is supported by an EPSRC Platform grant.

from: http://www.robots.ox.ac.uk/~vgg/data/oxbuildings/