2016 Imageclef Webupv Collection
Lists
Only you see this
Abstract
===============================================================================<br> Changelog:<br> 2018-05-05 Test data ground truth released.<br> <br> 2016-04-24 Test data for main subtask 3 is released. 2016-03-17 Test data for teaser tasks is released. 2016-03-16 A Bug was found in the Visual Feature files, please redownload if you use them 2016-02-17 Fixed the mixxing xml files in scaleconcept16_data_textual.webpages.tar.gz 2016-02-15 Added input document files for the development set of the teaser tasks.<br> * DevData/TeaserTasks/scaleconcept16.teaser_dev_input_documents.tar.gz 2016-02-15 Fixed some newline formatting issues in<br> * Features/scaleconcept16.teaser.TrainTestSplit.v20160215.tar.gz<br> Please download the latest version. 2016-02-12 Teaser development set: There were 2 duplicate webpages and 2<br> near-duplicate images in the previous release. Based on user<br> feedback, we have updated the dataset -- which now only includes<br> 3337 image-webpage pairs. Please download the latest versions of<br> the following to reflect these minor updates.<br> * DevData/TeaserTasks/scaleconcept16.teaser_dev_id.v20160212.tar.gz<br> * DevData/TeaserTasks/scaleconcept16.teaser_dev_data_textual.scofeat.v20160212.gz<br> * DevData/TeaserTasks/scaleconcept16.teaser2_dev_groundtruth.v20160212.txt<br> =============================================================================== This document describes the ScaleConcept dataset compiled for the ImageCLEF 2016<br> Scalable Concept Image Annotation challenge. The data mentioned here indicates<br> what is ready for download. However, upon request or depending on feedback<br> from the participants, additional data may be released. The following is the directory structure of the collection, and below there<br> is a brief description of what each compressed file contains. <br> Directory structure<br> ------------------- .<br> |<br> |--- README.txt<br> |--- scaleconcept16.agreement.txt.tar.gz<br> |--- scaleconcept16.concepts.tar.gz<br> |<br> |--- Features/<br> | |<br> | |--- scaleconcept16_ImgID.txt_Mod.tar.gz<br> | |--- scaleconcept16_ImgToTextID.tar.gz<br> | |--- scaleconcept16_TextID.txt_Mod.tar.gz<br> | |--- scaleconcept16.teaser.TrainTestSplit.*.tar.gz <br> | |<br> | |--- Textual/<br> | | |<br> | | |---scaleconcept16_data_textual.scofeat.tar.gz <br> | | |---scaleconcept16_data_textual.webpages.zip <br> | |<br> | |--- Visual/<br> | |<br> | |--- scaleconcept16_data_visual_gist.dfeat.gz<br> | |--- scaleconcept16_data_visual_sift_1000.sfeat.gz<br> | |--- scaleconcept16_data_visual_rgbsift_1000.sfeat.gz<br> | |--- scaleconcept16_data_visual_opponentsift_1000.sfeat.gz<br> | |--- scaleconcept16_data_visual_colorhist.sfeat.gz<br> | |--- scaleconcept16_data_visual_getlf.sfeat.gz<br> | |--- scaleconcept16_data_visual_vgg16-relu7.dfeat.gz<br> | |--- scaleconcept16_images.zip<br> |<br> |--- DevData/<br> | |<br> | |--- MainSubTasks/<br> | | |<br> | | |--- scaleconcept16.dev.visual.bbox.*.tar.gz<br> | | |--- scaleconcept16.dev.textdesc.*.tar.gz<br> | | |--- scaleconcept16.subtask3.dev.input_bbox.*.gz<br> | | |--- scaleconcept16.subtask3.dev.textdesc.*.gz<br> | |<br> | |--- TeaserTasks/<br> | | |<br> | | |--- scaleconcept16.teaser_dev_data_textual.scofeat.*.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_colorhist.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_csift_1000.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_getlf.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_gist.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_opponentsift_1000.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_rgbsift_1000.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_sift_1000.sfeat.gz<br> | | |--- scaleconcept16.teaser_dev_data_visual_vgg16-relu7.dfeat.gz<br> | | |--- scaleconcept16.teaser_dev_id.*.tar.gz<br> | | |--- scaleconcept16.teaser_dev_images.zip<br> | | |--- scaleconcept16.teaser_dev_pages.zip<br> | | |--- scaleconcept16.teaser2_dev_groundtruth.txt<br> |<br> |--- TestData/<br> | |<br> | |--- concepts.lst<br> | |--- scaleconcept16_subtask1_test.lst<br> | |--- scaleconcept16_subtask2_test.lst<br> | |--- scaleconcept16_subtask3_test.lst<br> | |--- scaleconcept16_subtask3_test.input_bbox.txt<br> | |--- scaleconcept16_teaser1_test_image_collection.lst<br> | |--- scaleconcept16_teaser1_test.lst<br> | |--- scaleconcept16_teaser2_test.lst<br> | |--- scaleconcept16_teaser_test_input_documents.tar.gz <br> Contents of files<br> ----------------- * scaleconcept16.concepts.tar.gz -> scaleconcept16.concepts.txt List of 251 concepts for the 2016 challenge. File format:<br> wordnet-offset \t category-word.pos.## \t list,of,synonyms,separated,by,commas \t defintiion -> scaleconcept16.concepts_hierarchy.txt The hierarchy structure of the 'general level' categories. File format:<br> *category \t *parent-category \t definition.<br> For example, *mammal is the child of *animal. '#' represents the root node. -> scaleconcept16.concepts_to_parents.txt List of 'general level' category parent(s) for each 251 concept. A concept may<br> have multiple parents (separated by commas). File format:<br> category \t *parent1,*parent2 * Features/scaleconcept16_ImgID.txt_Mod.tar.gz IDs of the images in the dataset. <br> * Features/scaleconcept16_TextID.txt_Mod.tar.gz IDs of the webpages in the dataset. <br> * Features/scaleconcept16_ImgToTextID.tar.gz IDs of images that appear on corresponding web pages <br> * Features/scaleconcept16.teaser.TrainTestSplit.*.tar.gz For Teasers 1 and 2: IDs of images and webpages, split into approximately<br> 300K for training and exactly 200K for testing.<br> The 200K test data cannot be explored during training for both teaser tasks. * Features/Textual/scaleconcept16_data_textual.scofeat.tar.gz The processed text extracted from the webpages near where the images<br> appeared. Each line corresponds to one image, having the same order<br> as the data_iids.txt list. The lines start with the image ID,<br> followed by the number of extracted unique words and the<br> corresponding word-score pairs. The scores were derived taking into<br> account 1) the term frequency (TF), 2) the document object model<br> (DOM) attributes, and 3) the word distance to the image. The scores<br> are all integers and for each image the sum of scores is always<br> <=100000 (i.e. it is normalized). <br> * Features/Textual/scaleconcept16_data_textual.webpages.tar.gz Contains all of the webpages which referenced the images in the<br> dataset set after being converted to valid xml. In total there are<br> 525766 files, since each image can appear in more than one page, and<br> there can be several versions of same page which differ by the<br> method of conversion to xml. To avoid having too many files in a<br> single directory (which is an issue for some types of partitions),<br> the files are found in subdirectories named using the first two<br> characters of the RID, thus the paths of the files after extraction<br> are of the form: ./scaleconcept16_data_textual.webpages/{RID:0:2}/{RID}.{CONVM}.xml.gz * Features/Visual/scaleconcept16_images.zip Contains thumbnails (maximum 640 pixels of either width or height)<br> of the images in jpeg format. To avoid having too many files in a<br> single directory (which is an issue for some types of partitions),<br> the files are found in subdirectories named using the first two<br> characters of the image ID, thus the paths of the files after extraction<br> are of the form: ./scaleconcepts16_images/{IID:0:2}/{IID}.jpg <br> * Features/Visual/scaleconcept16_*.{s|d}feat.gz The visual features in a simple ASCII text format either in sparse<br> (*.sfeat.gz files) or dense (*.dfeat.gz files). The first<br> line of the file indicates the number of vectors (N) and the<br> dimensionality (DIMS). Then each line corresponds to one vector.<br> For the dense features each line has exactly DIMS values separated<br> by spaces, i.e., the format is: N DIMS<br> Val(1,1) Val(1,2) ... Val(1,DIMS)<br> Val(2,1) Val(1,2) ... Val(2,DIMS)<br> ...<br> Val(N,1) Val(N,2) ... Val(N,DIMS) For the sparse features, each line starts with the number of non-zero<br> elements and is followed by dimension-value pairs, being the first<br> dimension 0, i.e., the format is: N DIMS<br> nz1 Dim(1,1) Val(1,1) ... Dim(1,nz1) Val(1,nz1)<br> nz2 Dim(2,1) Val(2,1) ... Dim(2,nz2) Val(2,nz2)<br> ...<br> nzN Dim(N,1) Val(N,1) ... Dim(N,nzN) Val(N,nzN) The order of the features is the same as in the list data_iids.txt. The procedure to extract the SIFT based features in this<br> subdirectory was conducted as follows. Using the ImageMagick<br> software, the images were first rescaled to having a maximum of 240<br> pixels, of both width and height, while preserving the original<br> aspect ratio, employing the command: convert {IMGIN}.jpg -resize '240>x240>' {IMGOUT}.jpg Then the SIFT features where extracted using the ColorDescriptor<br> software from Koen van de Sande<br> (http://koen.me/research/colordescriptors). As configuration we<br> used, 'densesampling' dete
BibTeX
@inproceedings{gilbert2016imageclef,
author = {Andrew Gilbert and Luca Piras and Josiah Wang and Fei Yan and Emmanuel Dellandréa and Robert Gaizauskas and Mauricio Villegas and Krystian Mikolajczyk},
title = {{2016 Imageclef Webupv Collection}},
booktitle = {Zenodo},
year = {2016},
doi = {10.5281/zenodo.1038554},
}Loading related content...