Home->Download->GOLD

GOLD




GOLD
Geo-Tagged Large Scale Dataset) contains more than 22K Flickr Crawled images together with their Geo-tags and it covers 80 different places of interest mined from a Flickr dataset which contains 3.3 million images. 

· This set of images is designed to be used as a test set for GPS location estimation and place of interest recognition.

· Besides the image, we also provide quantified code book and global features-color moment and HWVP.

· Along with the image we provide the location for each image.

· Dataset GOLD is partly focused on places of interest in China, Europe and America.

· GOLD dataset has been used for large scale image GPS location estimation in our TMM2013 paper and MMM2013 paper shown below.

Particularly the dataset includes images from Beijing, Shanghai, New York, Paris, London and Xi’an.(Images are in the Folder)




Dataset Rights
 

The GOLD dataset consists of images downloaded from Flickr, Baidu, and Google. Use of these images must respect the corresponding terms of use.

Citation

If you use the GOLD dataset please cite our papers:
[1]Jing Li, Xueming Qian, Yuan Yan Tang, Linjun Yang, Chaoteng Liu: GPS Estimation from Users' Photos. MMM (1) 2013: 118-129.
[2]Jing Li, Xueming Qian, Yuan Yan Tang, Linjun Yang, and Tao Mei, “GPS estimation for places of interest from social users’ uploaded photos”, IEEE Trans. Multimedia 2013.

Downloads

Small Images: A set of landmark and landscape images (22k images, 112M)
Global Features: Color Moment, HWVP (coming soon)
Local Features: Quantified Sift Code book (coming soon)

Publications

We list the papers using GOLD as follows:
[1]Jing Li, Xueming Qian, Yuan Yan Tang, Linjun Yang, Chaoteng Liu: GPS Estimation from Users' Photos. MMM (1) 2013: 118-129.
[2]Jing Li, Xueming Qian, Yuan Yan Tang, Linjun Yang, and Tao Mei, “GPS estimation for places of interest from social users’ uploaded photos”, IEEE Trans. Multimedia 2013.
[3]Yao Xue, Xueming Qian, and Baiqi Zhang “Mobile image retrieval using multi-photos as query”, In Proc. ICME, 2013.

 

 



 
ҳģҳģ