The large-scale eye-tracking database called LEDOV for video salinecy. Database from Lai Jiang, Mai Xu in Beihang University(2016).
LEDOV includes 538 videos with diverse content, containing a total of 179,336 frames and 6,431 seconds. The diverse content refers to the daily action, sports, social activity and art performance of human, and the videos of animal and man-man objects are also included. All videos are at least 720p resolution and 24 Hz frame rate. Then, for monitoring the binocular eye movements, a Tobii TX300 eye tracker was used in our experiment. Moreover, 32 participants (18 males and 14 females), aging from 20 to 56 (32 on average), were recruited to participate in the eye-tracking experiment. All participants were non-experts for the eye-tracking experiment, with normal/corrected-to-normal vision. During the experiment, the distance between subjects and the monitor was fixed at 65 cm. Before viewing videos, each subject was required to perform a 9-point calibration for the eye tracker. Afterwards, the subjects were asked to free-view videos displayed at random order. Finally, 5,058,178 fixations of all 32 subjects on 538 videos were collected for our eye-tracking database.
For the establishment details and data analysis of LEDOV, please refer to Section 3 of our paper.
Some examples of LEDOV are shown below and in 'sample.avi'.
All 538 videos as well as eye-tracking can be downloaded at Dropbox and BaiduYun
- 'VideoInfo.xlsx' and 'SubjectInfo.xlsx' list the information of videos and subjects.
- 'VideoInfo.mat' also includes the video information in the order of video name, frame rate, frame number, and resolution.
- 'Data.mat' in each sub-fold record the fixations of the video. In Data.fixdata, each row refers to a recorded fixation with the information as follows,
'column 1': the index of subject that the fixation belongs to
'column 2': the timestamp(ms) that the fixation starts
'column 3': the duration of the fixation
'column 4-5': the position of the fixation (take upper-lift corner as the origin)
As a test, you can run 'demo.m' to generate the heat-map series of the target video.
The evaluartion codes (AUC, NSS, CC and KL) we used in our paper can be refered in ./metrics. A example of using them can be seen in ./metrics/TestLEDOV.m
You are welcome to freely use this database, and please cite with the following Bibtex code:
@InProceedings{Jiang_2018_ECCV,
author = {Jiang, Lai and Xu, Mai and Liu, Tie and Qiao, Minglang and Wang, Zulin},
title = {DeepVS: A Deep Learning Based Video Saliency Prediction Approach},
booktitle = {The European Conference on Computer Vision (ECCV)},
month = {September},
year = {2018}
}
Should you have any queries, please contact jianglai.china@buaa.edu.cn / jianglai.china@aliyun.com