Emodb
| Domain | Emotions |
| Media | Image |
| Size | 4 GB tar-file |
| Instances | 1,27 million |
| File Format | JPEG |
| Creation Date | May, 2009 |
| Task | Classification |
| Copyright | http://diameter.itn.liu.se/emodb/ |
| URL | http://diameter.itn.liu.se/emodb/downloadform.php |
Description
brief description of data resource
The database contains 1.27 million thumbnails (maximum size 128 x 128 pixels) of images that were crawled from public web pages around March 2009. Images were crawled based on 70 different keywords related to emotions, such as "beautiful", "calm", "intense", etc. Saved thumbnails are accompanied by the URL to the original source, and metadata describing the popularity of each image.
Quality
format details (resolution, sampling rate, compression, ...)
Thumbnails (maximum size 128 x 128 pixels)
Source
source of data set and/or related event or project (e.g. TREC, FlickR, Corel, ...)
Public web pages
Ground Truth Annotation
is there an annotation or ground truth available? what kind of annotations? in what format (free text, structured, ..., file format)?
Thumbnails were shown to users visiting the Picsearch image search engine, and popularity measurements were derived based on the ratio between how many times each images has been viewed and clicked (only for images that have been viewed more than 50 times). The 100 most popular images in each category are marked as popular. The emodb database contains 70 image folders, corresponding to different categories/keywords. Each folder contains thumbnail images in jpeg format, together with a txt file of the same name as the folder.
Features
are there features (e.g. extracted from data content) available alongside the data set?
Each row in the txt file contains the following information (separated by a space) about the corresponding image:
ImageID URL Popularity
Where ImageID is a unique file name (without .jpg), URL is the url to the original source, and Popularity is set to 1 if the image is popular, otherwise 0. Thumbnail images are provided by ITN, Linköping University, whereas the metadata (user statistics) is provided by Picsearch AB (publ).
Copyright Remarks
remarks about copyright and usage restrictions
By downloading and using emodb, You agree to the following terms and conditions: 1. emodb can be used only for non-commercial research and educational purposes. 2. emodb is provided 'as-is', without any express or implied warranty. The authors or Linköping University can not be held liable for any direct or indirect damage arising out of downloading and using emodb. 3. The authors or Linköping University shall not be liable for any claim asserted against You by any other party, including but not limited to claims for infringement of copyright. 4. emodb may not be further distributed without permission from the authors, listed as contact persons at the bottom of this document.
Contact persons, responsible for emodb:
Martin Solli and Reiner Lenz ITN, Linköping University Sweden
{martin.solli, reiner.lenz} at liu.se
ITN, Campus Norrköping Linköping University SE-60174 Norrköping Sweden
Citation
references or publications
M. Solli, & R. Lenz, "Emotion Related Structures in Large Image Databases", ACM International Conference on Image and Video Retrieval, CIVR 2010, Xi'an, China, (2010).
External Links
link(s) to data set information and download
http://diameter.itn.liu.se/emodb/ (dataset information)
http://diameter.itn.liu.se/emodb/downloadform.php (download)
