=============================================================================== Changelog: 2018-05-05 Test data ground truth released. 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. * DevData/TeaserTasks/scaleconcept16.teaser_dev_input_documents.tar.gz 2016-02-15 Fixed some newline formatting issues in * Features/scaleconcept16.teaser.TrainTestSplit.v20160215.tar.gz Please download the latest version. 2016-02-12 Teaser development set: There were 2 duplicate webpages and 2 near-duplicate images in the previous release. Based on user feedback, we have updated the dataset -- which now only includes 3337 image-webpage pairs. Please download the latest versions of the following to reflect these minor updates. * DevData/TeaserTasks/scaleconcept16.teaser_dev_id.v20160212.tar.gz * DevData/TeaserTasks/scaleconcept16.teaser_dev_data_textual.scofeat.v20160212.gz * DevData/TeaserTasks/scaleconcept16.teaser2_dev_groundtruth.v20160212.txt =============================================================================== This document describes the ScaleConcept dataset compiled for the ImageCLEF 2016 Scalable Concept Image Annotation challenge. The data mentioned here indicates what is ready for download. However, upon request or depending on feedback from the participants, additional data may be released. The following is the directory structure of the collection, and below there is a brief description of what each compressed file contains. Directory structure ------------------- . | |--- README.txt |--- scaleconcept16.agreement.txt.tar.gz |--- scaleconcept16.concepts.tar.gz | |--- Features/ | | | |--- scaleconcept16_ImgID.txt_Mod.tar.gz | |--- scaleconcept16_ImgToTextID.tar.gz | |--- scaleconcept16_TextID.txt_Mod.tar.gz | |--- scaleconcept16.teaser.TrainTestSplit.*.tar.gz | | | |--- Textual/ | | | | | |---scaleconcept16_data_textual.scofeat.tar.gz | | |---scaleconcept16_data_textual.webpages.zip | | | |--- Visual/ | | | |--- scaleconcept16_data_visual_gist.dfeat.gz | |--- scaleconcept16_data_visual_sift_1000.sfeat.gz | |--- scaleconcept16_data_visual_rgbsift_1000.sfeat.gz | |--- scaleconcept16_data_visual_opponentsift_1000.sfeat.gz | |--- scaleconcept16_data_visual_colorhist.sfeat.gz | |--- scaleconcept16_data_visual_getlf.sfeat.gz | |--- scaleconcept16_data_visual_vgg16-relu7.dfeat.gz | |--- scaleconcept16_images.zip | |--- DevData/ | | | |--- MainSubTasks/ | | | | | |--- scaleconcept16.dev.visual.bbox.*.tar.gz | | |--- scaleconcept16.dev.textdesc.*.tar.gz | | |--- scaleconcept16.subtask3.dev.input_bbox.*.gz | | |--- scaleconcept16.subtask3.dev.textdesc.*.gz | | | |--- TeaserTasks/ | | | | | |--- scaleconcept16.teaser_dev_data_textual.scofeat.*.gz | | |--- scaleconcept16.teaser_dev_data_visual_colorhist.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_csift_1000.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_getlf.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_gist.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_opponentsift_1000.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_rgbsift_1000.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_sift_1000.sfeat.gz | | |--- scaleconcept16.teaser_dev_data_visual_vgg16-relu7.dfeat.gz | | |--- scaleconcept16.teaser_dev_id.*.tar.gz | | |--- scaleconcept16.teaser_dev_images.zip | | |--- scaleconcept16.teaser_dev_pages.zip | | |--- scaleconcept16.teaser2_dev_groundtruth.txt | |--- TestData/ | | | |--- concepts.lst | |--- scaleconcept16_subtask1_test.lst | |--- scaleconcept16_subtask2_test.lst | |--- scaleconcept16_subtask3_test.lst | |--- scaleconcept16_subtask3_test.input_bbox.txt | |--- scaleconcept16_teaser1_test_image_collection.lst | |--- scaleconcept16_teaser1_test.lst | |--- scaleconcept16_teaser2_test.lst | |--- scaleconcept16_teaser_test_input_documents.tar.gz Contents of files ----------------- * scaleconcept16.concepts.tar.gz -> scaleconcept16.concepts.txt List of 251 concepts for the 2016 challenge. File format: 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: *category \t *parent-category \t definition. 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 have multiple parents (separated by commas). File format: category \t *parent1,*parent2 * Features/scaleconcept16_ImgID.txt_Mod.tar.gz IDs of the images in the dataset. * Features/scaleconcept16_TextID.txt_Mod.tar.gz IDs of the webpages in the dataset. * Features/scaleconcept16_ImgToTextID.tar.gz IDs of images that appear on corresponding web pages * Features/scaleconcept16.teaser.TrainTestSplit.*.tar.gz For Teasers 1 and 2: IDs of images and webpages, split into approximately 300K for training and exactly 200K for testing. 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 appeared. Each line corresponds to one image, having the same order as the data_iids.txt list. The lines start with the image ID, followed by the number of extracted unique words and the corresponding word-score pairs. The scores were derived taking into account 1) the term frequency (TF), 2) the document object model (DOM) attributes, and 3) the word distance to the image. The scores are all integers and for each image the sum of scores is always <=100000 (i.e. it is normalized). * Features/Textual/scaleconcept16_data_textual.webpages.tar.gz Contains all of the webpages which referenced the images in the dataset set after being converted to valid xml. In total there are 525766 files, since each image can appear in more than one page, and there can be several versions of same page which differ by the method of conversion to xml. To avoid having too many files in a single directory (which is an issue for some types of partitions), the files are found in subdirectories named using the first two characters of the RID, thus the paths of the files after extraction 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) of the images in jpeg format. To avoid having too many files in a single directory (which is an issue for some types of partitions), the files are found in subdirectories named using the first two characters of the image ID, thus the paths of the files after extraction are of the form: ./scaleconcepts16_images/{IID:0:2}/{IID}.jpg * Features/Visual/scaleconcept16_*.{s|d}feat.gz The visual features in a simple ASCII text format either in sparse (*.sfeat.gz files) or dense (*.dfeat.gz files). The first line of the file indicates the number of vectors (N) and the dimensionality (DIMS). Then each line corresponds to one vector. For the dense features each line has exactly DIMS values separated by spaces, i.e., the format is: N DIMS Val(1,1) Val(1,2) ... Val(1,DIMS) Val(2,1) Val(1,2) ... Val(2,DIMS) ... Val(N,1) Val(N,2) ... Val(N,DIMS) For the sparse features, each line starts with the number of non-zero elements and is followed by dimension-value pairs, being the first dimension 0, i.e., the format is: N DIMS nz1 Dim(1,1) Val(1,1) ... Dim(1,nz1) Val(1,nz1) nz2 Dim(2,1) Val(2,1) ... Dim(2,nz2) Val(2,nz2) ... 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 subdirectory was conducted as follows. Using the ImageMagick software, the images were first rescaled to having a maximum of 240 pixels, of both width and height, while preserving the original aspect ratio, employing the command: convert {IMGIN}.jpg -resize '240>x240>' {IMGOUT}.jpg Then the SIFT features where extracted using the ColorDescriptor software from Koen van de Sande (http://koen.me/research/colordescriptors). As configuration we used, 'densesampling' detector with default parameters, and a hard assignment codebook using a spatial pyramid as 'pyramid-1x1-2x2'. The number in the file name indicates the size of the codebook. All of the vectors of the spatial pyramid are given in the same line, thus keeping only the first 1/5th of the dimensions would be like not using the spatial pyramid. The codebook was generated using 1.25 million randomly selected features and the k-means algorithm. The GIST features were extracted using the LabelMe Toolbox. The images where first resized to 256x256 ignoring original aspect ratio, using 5 scales, 6 orientations and 4 blocks. The other features colorhist and getlf, are both color histogram based extracted using our own implementation. * Features/Visual/scaleconcept16_data_visual_vgg16-relu7.dfeat.gz Contains the 4096 dimensional activations of the relu7 layer of Oxford VGG's 16-layer CNN model, extracted using the Berkeley Caffe library. More details can be found at https://github.com/BVLC/caffe/wiki/Model-Zoo. * DevData/MainSubTasks/scaleconcept16.dev.viusal.bbox.*.tar.gz Development set ground truth localised annotations for sub task 1. The format for the development set of annotated bounding boxes of the concepts is <image_ID> <seq> <Concept> <confidence> <xmin> <ymin> <xmax> <ymax> The development set contains 1,979 images. The bounding boxes may enclose single instances (a single tree) or grouped instances (e.g. group of trees), depending on the context. The annotations are not exhaustive: the emphasis is on concepts that are interesting enough to be described in the image, although background objects are also optionally annotated by our annotators in many cases. Also note that a person might not be annotated if the annotator could not decide whether the person is a man/woman/boy/girl. * DevData/MainSubTasks/scaleconcept16.dev.textdesc.*.tar.gz Development set ground truth textual description annotations of images for Subtask 2 The format is: <image_ID> \t <text_description_seq> \t <textual_description> The development set contains 2,000 images with 5 to 51 textual descriptions per image (mean: 9.492, median: 8). Please note that the sentences contain a mix of both American and British English spelling variants (e.g. color vs colour) -- we have decided to retain this variation in the annotations to reflect the challenge of real-world English spelling variants. Basic spell-correction has been performed on the textual descriptions, but we cannot guarantee that they are completely free from spelling or grammatical error. * DevData/MainSubTasks/scaleconcept16.subtask3.dev.input_bbox.*.gz Input bounding boxes for Subtask 3. This is a selected subset of 500 development images from scaleconcept16.dev.visual.bbox above (please refer to above for file format). * DevData/MainSubTasks/scaleconcept16.subtask3.dev.textdesc.*.gz Annotated textual descriptions for 500 development images, to be used to evaluate the content selection ability of the text generation system in the clean track of SubTask 3. The format is the same as the original scaleconcept16.dev.textdesc file, except that we further annotated textual terms with their corresponding input bounding boxes, for example [[[dogs|0,4]]] in a textual description refers to the two instances of dogs with the bounding box id 0 and 4 in scaleconcept16.subtask3.dev.input_bbox. Note that not all descriptions from the original scaleconcept16.dev.textdesc are used in this version, and as such the sequence numbers of the descriptions may not necessarily be contiguous as we retained the sequence numbers from the original file for consistency. * DevData/TeaserTasks/scaleconcept16.teaser_dev_id.*.tar.gz -> scaleconcept16.teaser_dev.ImgID.txt IDs of 3339 images for the development set of both teaser tasks. Note that 2 images are near-duplicates and will thus not be used in this dataset. We have left them intact to avoid having participants re-download the visual features. -> scaleconcept16.teaser_dev.TextID.txt IDs of 3337 webpage documents for the development set of both teaser tasks. -> scaleconcept16.teaser_dev.ImgToTextID.txt IDs of 3337 images that appear on corresponding web pages. * DevData/TeaserTasks/scaleconcept16.teaser_dev_input_documents.tar.gz -> scaleconcept16.teaser_dev.docID.txt IDs of 3337 input text documents for the development set of both teaser tasks. -> docs/{docID} The text for each input document. These should be used as input for both teaser tasks. * DevData/TeaserTasks/scaleconcept16.teaser_dev_images.zip Contains 3339 images for the development set of both teaser tasks in jpeg format. * DevData/TeaserTasks/scaleconcept16.teaser_dev_pages.zip Contains 3337 webpages for the development set of both teaser tasks after converting to valid xml format, each compressed as a gzip file. * DevData/TeaserTasks/scaleconcept16.teaser_dev_data_textual.scofeat.*.gz The processed text extracted from 3337 webpages near where the images appeared. Please refer to Features/Textual/scaleconcept16_data_textual.scofeat.tar.gz above for more details. * DevData/TeaserTasks/scaleconcept16_teaser_dev_data_visual_*.{s|d}feat.gz The visual features for the 3339 images for the development set of both teaser tasks. Please refer to Features/Visual/scaleconcept16_*.{s|d}feat.gz above for more details. * DevData/TeaserTasks/scaleconcept16.teaser2_dev_groundtruth.*.txt The GPS coordinates for 3337 documents from the development set for Teaser Task 2 (Geolocation). File format: WebpageID latitude longitude * TestData/concepts.lst List of 251 concepts for the main subtasks 1, 2, and 3. File format: wordnet-offset \t category-word.pos.## * TestData/scaleconcept16_subtask{1|2}_test.lst List of 510,123 images to annotate for main subtasks 1 and 2. Both files are identical. * TestData/scaleconcept16_subtask3_test.lst List of 450 images to annotate for main subtask 3. * TestData/scaleconcept16_subtask3_test.input_bbox.txt List of bounding boxes for 450 test images, to be used as input for main subtask 3. The format is the same as the development set. * TestData/scaleconcept16_teaser1_test_image_collection.lst The collection of 200,000 test images to be used for Teaser Task 1 (text illustration) * TestData/scaleconcept16_teaser1_test.lst The list of IDs for 180,000 text documents to be used as input for Teaser Task 1. Please note that the IDs are *not* the same as the webpage IDs provided in the 500K corpus. The task is to provide a ranked list of the top 100 images (from the 200,000 test image collection above) for each input text document. * TestData/scaleconcept16_teaser2_test.lst The list of IDs for 180,000 text documents to be used as input for Teaser Task 2 (identical to teaser task 1). The task is to provide the latitude and longitude for each input text document. * TestData/scaleconcept16_teaser_test_input_documents.tar.gz The input text documents to be used for Teaser Tasks 1 and 2. These should be used as input for both teaser tasks. * TestData/scaleconcept16_groundtruth.zip Ground truth for the test set. Contact ------- For further questions, please contact: Andrew Gilbert <a.gilbert@surrey.ac.uk>