SFM MODELS FOR YFCC100M DATASET =============================== The structure-from-motion models included in this package are licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License: http://creativecommons.org/licenses/by-nc-sa/4.0/ The structure-from-motion models included in this package were generated using the method described in the following publication: @inproceedings{heinly2015_reconstructing_the_world, title = {{Reconstructing the World* in Six Days *(As Captured by the Yahoo 100 Million Image Dataset)}}, author = {Jared Heinly and Johannes Lutz Sch\"{o}nberger and Enrique Dunn and Jan-Michael Frahm}, booktitle = {Computer Vision and Pattern Recognition (CVPR)}, year = {2015} } The YFCC100M dataset was obtained from Yahoo: http://labs.yahoo.com/news/yfcc100m/ http://webscope.sandbox.yahoo.com/catalog.php?datatype=i&did=67 Please cite both the above publication as well as the Yahoo dataset if you use these models in your own work. The models are stored in the NVM file format, which is used by VisualSFM: http://ccwu.me/vsfm/ http://ccwu.me/vsfm/doc.html#nvm To determine which images were used in the reconstructions, refer to the filenames in the NVM file (model.nvm). Each filename has the form 12345678_0123456789.jpg, where 12345678 is a unique 8-digit identifier used by our system, and 0123456789 is the unique Flickr photo_id of the image. Given the Flickr photo_id, you can query the Flickr API to learn more about the photo: https://www.flickr.com/services/api/flickr.photos.getInfo.html https://www.flickr.com/services/api/flickr.photos.getExif.html Each model folder contains a list of the Flickr users whose photos were used in the reconstruction (attribution.txt). Each model folder also contains a list of the image dimensions for the photos that were used in the reconstruction (images.dim).