Estás aquí: start » aws-d » citius_video_database

INDEX

Esta é unha revisión antiga do documento!


INDEX

CITIUS Video Database - Information

This eye tracking video database can be used to validate visual attention models. This dataset includes 72 videos downloaded from Internet and some synthetic videos generated in the lab. The videos can be classified in four categories, natural and synthetic, with fixed or movement camera. It includes 27 synthetic videos with dynamic pop-out effects. The videos have been selected in order to minimize the influence of the top-down efects.

Videos

  • Number of videos … 72
  • Video resolution ….. 320×240 pixels
  • Frame rate (fps) ….. 15 fps (Natural) / 30 fps (Synthetic)
  • Video format ………. Mpeg 4
  • Video duration …….. Around 10 seconds. (198±93 Frames per video).

Subjective test

  • Observation distance .. The head of the subject was placed on a chin-rest located at 60cm from the display.
  • Environment ………….. ITU-R BT.500-11
  • Duration ………………. 11,4 min.
  • Free or Quality Task … Its a Task Free Experiment.
  • Calibration procedure .. The calibration procedure was carried out at the beginning of each block. The subject had to follow a target that was placed sequentially at 5 locations evenly distributed across the screen. Accuracy of the calibration was then validated at 4 of these calibrated locations. If the error in the estimated position was greater than 0.5 degrees of visual angle, the experiment was stopped and calibration restarted.

Eyetracker

  • Eyetracker …………….. SMI RED Eyetracker
  • Eyetracker mode ……… Binocular
  • Acquisition frequency … 50Hz

Observers

  • Number of observers … 22 naives with ages ranging from 11 to 43 (All the videos were presented in random order to each subject).
  • Number of fixations ….. 40.558 (Left eye) / 40.348 (Right eye) Extracted with BeGaze™ Software.
  • Subject-Video hours …. 4,19h (22 subjects*11,4min)

Display

  • Display model …….. The LCD display had a 1280×1024 pixels resolution, with a physical screen size of 33,792mm.(H) x 27,0336mm.(V)
  • Display resolution … 1280×1024
  • Display luminance … 0.65/450 cd/m²
  • Display frequency …. 60Hz

Fixation database structure

Each video has been reescaled to a 320×240 resolution and the fixation information for all the subjects is refered to the same image space. X and Y are image's coordinates refered to the leftmost bottom corner, and both parameters Start and Duration are in ms. All the fixations have been saved in the CITIUS_VDB_Fixations file,CITIUS_VDB_Fixations, using matlab with the following structure:

EyeTrackVDB
|-Video (List of Videos)
   → Name (Original video name)
   → ShortName (Used in publications)
   → NF   (Number of frames)
   → FPS  (Frames per second)
   → FixL (List of fixations for the left eye:  [X,Y,Start,Duration])
   → FixR (List of fixations for the right eye: [X,Y,Start,Duration])
   → Frame (List of Frames)
      |-Time (Start time for each frame)
      |-FixL (List of left fixations that happened during this frame: [X,Y,Start,Duration])
      |-FixR (List of right fixations that happened during this frame: [X,Y,Start,Duration])  

How can I use this fixation database structure?

In the following example we illustrate how to access to the XY coordinates of the left eye fixations in the seventh frame of the third video of the database:

>> load('EyeTrackVDB.mat');

>> disp(EyeTrackVDB.Video(3).Name);

RCD_HelicopterCar_2.avi

>> disp(EyeTrackVDB.Video(3).Frame(7).FixL(:,1:2));

 184   125
 229   156
 178   140
  95   136
  95   108
 229   174
 ...   ...
 220   154
 228    98
 143   121

MATLAB script for the visualization of all the fixations

The MATLAB script “WEB_ShowVideoAndFixations.m” is another example that reads the MATLAB database file. This function shows all the fixations, for all the subjects, sumperimposed to the original video frames. It only needs the path of the database file and the name of the selected video. The stand alone database compressed file “CITIUS_VDB” includes all the folders with the needed information. The included MATLAB script assumes the following folder structure:

  • The folder “.\DataBases\Scripts\” — — — — — — — — — — - —→ should include this “.m” file.
  • The folder “.\DataBases\CITIUS\Videos\” — — — — — — — - - —→ should include the video files.
  • The folder “.\DataBases\CITIUS\HumanData\EyeTrackData\” - - —→ should include the “EyeTrackVDB.mat” file with the database information.

For example, to use this script to show the video “RCD_Surf_2.avi” with the fixations superimposed you should open matlab, change to the Scripts folder and type the following sentence at the command line. If you change, inside the code, the value of the boolean variable bVisFixPOS, you can see the fixation positions superimposed or not.

>> WEB_ShowVideoAndFixations('..\DataBases\CITIUS\HumanData\eyeTrackData\EyeTrackVDB.mat','RCD_Surf_2');

Original Video Video with Saliency Map Video with Map and Fixations
RCD_Surf_2 bVisFixPOS=false bVisFixPOS=true

or, if you want to show all the videos of the CITIUS database, in a new MATLAB figure, you can type:

>> WEB_ShowVideoAndFixations('..\DataBases\CITIUS\HumanData\eyeTrackData\EyeTrackVDB.mat',[]);

You can also use filters, for instance to show all the videos that contain the string 'RCD' in the name use:

>> WEB_ShowVideoAndFixations('..\DataBases\CITIUS\HumanData\eyeTrackData\EyeTrackVDB.mat','RCD');

Thumbnails of videos

  • The following images show thumbnails for the Natural videos in the CITIUS database. All the videos included on the database have been rescaled to a resolution of 320×240 pixels.
Real with Static CameraNatural with Static Camera
autoplay=1|rce_berlinalexanderplatz_2 autoplay=1|rce_berlinalexanderplatz_4 autoplay=1|rce_door_1 autoplay=1|rce_fish_1 autoplay=1|rce_longskate_2 autoplay=1|rce_longskate_3 autoplay=1|rce_lucky_1
autoplay=1|rce_parking_1 autoplay=1|rce_parking_2 autoplay=1|rce_parking_3 autoplay=1|rce_parkingnight_1 autoplay=1|rce_rain_1 autoplay=1|rce_snowtrain_1 autoplay=1|rce_streetdance_3
autoplay=1|rce_traffic_1 autoplay=1|rce_traffic_2 autoplay=1|rce_train_1
Real with Dynamic CameraNatural with Dynamic Camera
autoplay=1|rcd_basejumping_1 autoplay=1|rcd_eagle_1 autoplay=1|rcd_extremeski_1 autoplay=1|rcd_extremeski_2 autoplay=1|rcd_extremeski_3 autoplay=1|rcd_extremeski_4 autoplay=1|rcd_extremeski_5
autoplay=1|rcd_extremeski_6 autoplay=1|rcd_helicoptercar_1 autoplay=1|rcd_helicoptercar_2 autoplay=1|rcd_helicoptercar_3 autoplay=1|rcd_iceskate_1 autoplay=1|rcd_lion_1 autoplay=1|rcd_lion_2
autoplay=1|rcd_lion_3 autoplay=1|rcd_lion_4 autoplay=1|rcd_lion_5 autoplay=1|rcd_longskate_1 autoplay=1|rcd_surf_1 autoplay=1|rcd_surf_2 autoplay=1|rcd_surf_3
autoplay=1|rcd_surf_4 autoplay=1|rcd_surf_5 autoplay=1|rcd_surf_6 autoplay=1|rcd_surf_7 autoplay=1|rcd_tuna_1 autoplay=1|rcd_tuna_2 autoplay=1|rcd_tuna_3
  • The following images show thumbnails for the Synthetic videos in the CITIUS database.
Synthetic with Static CameraSynthetic with Static Camera
autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1|
autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1|
autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1| autoplay=1|
autoplay=1| autoplay=1|
Synthetic with Dynamic CameraSynthetic with Dynamic Camera
autoplay=1| autoplay=1| autoplay=1| autoplay=1|

Download All

If you only want to download the videos you can do it by clicking on the download icons of the previous section. If you have a fast enough internet connection and you prefer to download a stand alone file, with all the information and folder structure included inside, you can download by clicking on this link:

CITIUS_VDB_Videos

Research Team

Bib reference

The database, along with MATLAB functions for its use, will be freely downloaded here, and will be used without restriction for educational and research purposes, providing that our paper “Dynamic Whitening Saliency” from IEEE Transactions on Pattern Analysis and Machine Intelligence ( t-PAMI), is cited in any published work.

@article{10.1109/TPAMI.2016.2567391,
author = {Leborán Alvarez, Víctor and García-Díaz, Antón and Fdez-Vidal, Xosé R. and Pardo, Xosé M.},
title = {Dynamic Whitening Saliency},
journal ={IEEE Transactions on Pattern Analysis and Machine Intelligence},
volume = {XX},
number = {XX},
issn = {0162-8828},
year = {2016},
pages = {XXXX-XXXX},
doi = {http://doi.ieeecomputersociety.org/10.1109/TPAMI.2016.2567391},
publisher = {IEEE Computer Society},
address = {Los Alamitos, CA, USA},
}

Other video databases

  • CRCNS - Collaborative Research in Computational Neuroscience : 50 videos viewed by 8 users (L. Itti 2006).
  • IRCCyN IVC Eyetracker SD 2008 11 : 51 videos viewed by 37 users (Fadi Boulos, Wei Chen, Benoit Parrein and Patrick Le Callet 2008).
  • The DIEM Project: 85 videos viewed by 250 users (John M. Henderson, Robin Hill,Tim Smith and K. Mital 2009).
  • Lubeck University Dataset: 54 videos viewed by 18 users (Michael Dorr, Thomas Martinetz, Karl Gegenfurtner and Erhardt Barth 2010).
  • Actions in the Eye: 497107 frames viewed by 16 users (Stefan Mathe and Cristian Sminchisescu 2012).
  • Eye tracking database for standard video sequences: 12 Videos viewed by 15 users (H. Hadizadeh, M. J. Enriquez, and I. V. Bajic 2012)
  • ASCMN database: The database contains 24 videos split into 5 classes (N. Riche, M. Mancas, D. Kulibrk, V. Krnojevic, B. Gosselin, T. Dutoit).
  • Hollywood2 : It includes 69 movies used to generate the 823 train and 884 test clips in this dataset (Marcin Marszalek and Ivan Laptev and Cordelia Schmid).

License

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. This information is under the license Creative Commons Reconocimiento-Compartir Igual 4.0 Internacional>. You can use this dataset on your publication as long as you include a citation to the reference on this page. When including a link to this dataset, please use this page instead of linking the file directly.

Click here to return the Top of this page …