Semantic Image-Text-Classes

This dataset is introduced by the paper "Understanding, Categorizing and Predicting Semantic Image-Text Relations". If you are using this dataset it in your work, please cite: ``` @inproceedings{otto2019understanding, title={Understanding, Categorizing and Predicting Semantic Image-Text Relations}, author={Otto, Christian and Springstein, Matthias and Anand, Avishek and Ewerth, Ralph}, booktitle={In Proceedings of ACM International Conference on Multimedia Retrieval (ICMR 2019)}, year={2019} } ``` To create the full tar use the following command in the command line: > cat train.tar.part* > train_concat.tar Then simply untar it via > tar -xf train_concat.tar The jsonl files contain metadata of the following format: > id, origin, CMI, SC, STAT, ITClass, text, tagged text, image_path License Information: This dataset is composed of various open access sources as described in the paper. We thank all the original authors for their work. * Pitt Image Ads Dataset: http://people.cs.pitt.edu/~kovashka/ads/ * Image-Net challenge: http://image-net.org/ * Visual Storytelling Dataset (VIST): http://visionandlanguage.net/VIST/ * Wikipedia: https://www.wikipedia.org/ * Microsoft COCO: http://cocodataset.org/#home