Copyright © CSP Lab. Communications & Signal Processing Laboratory, Dept Information Engineering, University of Florence, 2021.
This work is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.
PREMIER Dataset v1 was supported in part by the Italian Ministry of Education, Universities and Research MIUR under Grant 2017Z595XS, and in part by DARPA under Grant FA8750-16-2-0188.
PREMIER Dataset v1 is available at Drive-Link.
The dataset is characterized by a family of native data marked by the suffix -Nx and an altered one marked by the suffix -Ax. This structure allows having consistent data but also the necessary flexibility to expand and represent future technologies.
The contents of the PREMIER-Dataset folder is:
For each dataset component, several CSV files are provided: videos/images details and a summary. The videos/images specifics are marked by the suffix -videos or -images. For example, the additional information given for PREMIER-N1 follow:
PREMIER-N1 contains 8 smartphones: a Samsung Galaxy S7, a Lenovo P2, a Wiko PULP FAB 4G, an Apple iPhone SE, a Huawei P10 Lite, an Apple iPhone 3GS, an Apple iPhone 4S, and an Apple iPhone 5S. The collection includes 26 videos of flat/indoor/outdoor scenery and 987 flat/natural images.
For each CSV file is used the following schema:
ID|Brand|Model|Firmware|nVideos|nImagesID|Filename|Brand|Model|Scene|Compression|Image Height|Image WidthID|Filename|Brand|Model|Scene|Major Brand|Compressor ID|Video Height|Video Width|Rotation|DurationThe elements of the schema are described as follows:
PREMIER-N2 contains 5 devices: a Xiaomi Redme Note 8T, a Google Pixel 3a, a Motorola Moto G9 Plus, a Huawei P10 Lite and an Apple iPhone XS. The image collection includes flat and natural scenes, when available also RAW and HEIC formats are included. The video collection contains flat, indoor and outdoor scenery with and without movement. Some additional videos are provided to evaluate H265/HEVC codec and other non-default resolutions.
For each CSV file is used the following schema:
ID|Brand|Main Model|Alt. Model|Main Firmware|Alt. Firmware|nVideos|nImagesID|Filename|Brand|Main Model|Alt. Model|Main Firmware|Alt. Firmware|Scene|Compression|Image Height|Image WidthID|Filename|Brand|Main Model|Alt. Model|Main Firmware|Alt. Firmware|Scene|Major Brand|Compressor ID|Video Height|Video Width|Rotation|DurationThe elements of the above schema are the same as in the PREMIER-N1 description. The following elements differ:
An additional file is included in this dataset, PREMIER-N2-log-rename.csv. It contains for each media its original name and the one used by PREMIER-N2. This file may be useful in order to keep trace of meta information that some devices include in the naming convention of the file.
The PREMIER-N2-log-rename.csv schema is Device|Sub-folders|Original name|PREMIER name.
The elements of the schema are described as follows:
PREMIER-A1 is the edited subset of EVA-7K (available at https://lesc.dinfo.unifi.it/en/datasets).
PREMIER-A1 contains 1400 videos edited via Avidemux, Ffmpeg, Kdenlive, and Adobe Premiere. The editing processes include speed up, slow down, frame deletion and re-encoding.
Furthermore, each edited video is also shared through YouTube, Facebook, Weibo, and TikTok social platforms.
For each CSV file is used the following schema:
ID|Brand|Model|Firmware|non-SN|Facebook|Youtube|Tiktok|WeiboID|Filename|Brand|Model|Origin|Editing|Major Brand|Compressor ID|Video Height|Video Width|Rotation|DurationThe elements of the above schema are the same as in the PREMIER-N1 description. The following elements differ:
PREMIER-A2 contains videos shared through Facebook and Youtube social networks.
40 native videos selected from the VISION dataset have been single shared through Facebook and Youtube.
40 videos are double shared via Facebook-Youtube.
40 videos are double shared via Youtube-Facebook sharing chains.
For each CSV file is used the following schema:
ID|Brand|Model|Firmware|EVA-Native|Facebook|Youtube|Facebook-Youtube|Youtube-FacebookID|Filename|Brand|Model|Origin|Major Brand|Compressor ID|Video Height|Video Width|Rotation|DurationThe elements of the above schema are the same as in the PREMIER-N1 description. The following elements differ:
PREMIER-A3 contains videos selected from 20 devices of Video-ACID and NYUAD-MMD datasets.
The original 80 videos have been exchanged through Facebook, Instagram, Telegram, Twitter, and YouTube social platforms.
For each CSV file is used the following schema:
ID|Brand|Model|Firmware|Native Dataset|Facebook|Instagram|Telegram|Twitter|YoutubeID|Filename|Brand|Model|Origin|Major Brand|Compressor ID|Video Height|Video Width|Rotation|DurationThe elements of the above schema are the same as in the PREMIER-N1 description. The following elements differ:
todo