Datasets





 Note: The following links are available only via Mozilla Firefox as from 25/03/2020.
 
  
 The dataset consists of outdoor content acquired from 46 smartphones of 11 major brands. For each device, we collected a total of 9206 media contents (6637 images and 1831 videos). The natural images and videos captured with each smartphone portray the same subjects, in order to reduce biases that could influence the performance of the analysis methods.
The complete dataset will be released as soon as the review process of the corresponding manuscript is finalised.  
 
 
 The 'EVA-7K dataset' contains 7000 videos: native, altered and exchanged through social platforms. The altered contents include manipulations with FFmpeg, AVIdemux, Kdenlive and Adobe Premiere. The social platforms used to exchange the native and altered videos are Facebook, Tiktok, Youtube and Weibo. A detailed description of the dataset is available in the journal paper "Efficient Video Integrity Analysis Through Container Characterization", published on IEEE Journal of Selected Topics in Signal Processing, 2020.
 
 
 
 The 'HDR dataset' contains more than 5000 Standard Dynamic Range (SDR) and High Dynamic Range (HDR) images captured using 23 different mobile devices of 7 major brands. A detailed description of the dataset is available in the journal paper "A New Dataset for Source Identification of High Dynamic Range Images", published on Sensors, 2018.
 
If you download and use contents of the 'HDR dataset' you agree to the following HDR license agreement. We are pleased to receive electronic copies of any publication making use of our database and to add your reference to the list of related publications.
 
 
 
 The 'VISION dataset' contains more than 35000 images and videos captured using 35 different portable devices of 11 major brands. A detailed description of the dataset is available in the scientific, open acess paper "VISION: a video and image dataset for source identification", published on EURASIP Journal on Information Security on Dec. 2017, freely available at this url: https://link.springer.com/content/pdf/10.1186%2Fs13635-017-0067-2.pdf.
 
If you download and use contents of the 'VISION dataset' you agree to the following VISION license agreement. We are pleased to receive electronic copies of any publication making use of our database and to add your reference to the list of related publications.
 
ERRATA -- 17/04/2019
 
We regret to inform the VISION users that the following videos have been misplaced therefore we suggest to not consider them as native (and social) contents in your analysis.
- D03_V_flat_still_0002.mp4, D03_V_flatYT_still_0002.mp4, D03_V_flatWA_still_0002.mp4
- D19_V_flat_move_0002.mov, D19_V_flatYT_move_0002.mp4, D19_V_flatWA_move_0002.mp4

 
 
 
Dataset used in the experiments reported in T.Bianchi, A.Piva, "Image Forgery Localization via Block-Grained Analysis of JPEG Artifacts",  IEEE Transactions on Information Forensics & Security, vol. 7,  no. 3,  June 2012,  pp. 1003 - 1017.
 
Creative Common
Image dataset for localization of double JPEG compression by Alessandro Nozzoli, Università di Firenze is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
 
 
 
 Dataset used in the experiments reported in Albisani, Chiara, Massimo Iuliani, and Alessandro Piva. "Checking PRNU Usability on Modern Devices." ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021.