Video Saliency Detection is the process of detecting or predicting the most visually important or attention-grabbing regions in video sequence. This regions are where humans most likely focus their attention when watching the video.
데이터셋
출처에 같은 태스크 이름으로 연결된 데이터셋입니다. 동명 태스크가 포함될 수 있으므로, 비교할 때는 평가 결과의 조건도 확인하세요.
DHF1K is a video saliency dataset which contains a ground-truth map of binary pixel-wise gaze fixation points and a continuous map of the fixation points after being blurred by a gaussian filter. DHF1K contains 1000 videos in total. 700 of the videos are annotated, 600 of which are used for training and 100 for validation. The remaining 300 are the testing set which are to be evaluated on a public server. Source: ViP: Video Platform for PyTorch Image Source: https://arxiv.org/pdf/1801.07424.pdf
The dataset presents open high-resolution test clips set with different types of content: movie fragments, sport streams, live caption clips. Used clips of 1920×1080 resolution and with duration from 13 to 38 seconds. And Performed reliable data collection from 50 observers (19–24 y. o.) using 500 Hz SMI iViewXTM Hi-Speed 1250 eye-tracker. Also used cross-fade which ensures the independence of the received fixations between different clips. The final ground-truth saliency map was estimated as a Gaussian mixture with centers at the fixation points. A standard deviation for the Gaussians equal to 120 was chosen (this value matches 8 angular degrees, which is known to be the sector of sharp vision).
A novel audio-visual mouse saliency (AViMoS) dataset with the following key-features: * Diverse content: movie, sports, live, vertical videos, etc.; * Large scale: 1500 videos with mean 19s duration; * High resolution: all streams are FullHD; * Audio track saved and played to observers; * Mouse fixations from >5000 observers (>70 per video); * License: CC-BY;