/video2gif_dataset

The Video2GIF dataset with 100k GIFs from our paper at CVPR2016

Primary LanguageShellBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Please note: We provide the pre-trained model here: https://github.com/gyglim/video2gif_code

================================================================================

Video2GIF dataset, version 0.9

The Video2GIF dataset contains over 100,000 pairs of GIFs and their source 
videos. The GIFs were collected from two popular GIF websites (makeagif.com, 
gifsoup.com) and the corresponding source videos were collected from YouTube in 
Summer 2015. We provide IDs and URLs of the GIFs and the videos, along with 
temporal alignment of GIF segments to their source videos. The dataset shall be 
used to evaluate GIF creation and video highlight techniques.

In addition to the 100K GIF-video pairs, the dataset contains 357 pairs of GIFs 
and their source videos as the test set. The 357 videos come with a Creative 
Commons CC-BY license, which allows us to redistribute the material with 
appropriate credit. We provide this test set to make the results reproducible 
even when some of the videos become unavailable.

If you end up using the dataset, we ask you to cite the following paper:

    Michael Gygli, Yale Song, Liangliang Cao
    "Video2GIF: Automatic Generation of Animated GIFs from Video,"
    IEEE CVPR 2016

If you have any question regarding the dataset, please contact:

    Michael Gygli <gygli@vision.ee.ethz.ch>

License: This dataset is licensed under BSD, see LICENSE file

================================================================================

Full description:

This repo has the following content:

(a) "metadata.txt" 
    This file contains information about GIFs, their corresponding source video 
    ID, and temporal alignment of the GIF to the video. It contains the following 
    fields:

    youtube_id, is_creative_commons, gif_id, gif_url, gif_start_frame, 
    gif_end_frame, gif_start_sec, gif_end_sec, gif_title, gif_views, gif_age, 
    video_title, video_views, video_publish_date, video_likes, video_duration, 
    video_frame_count, video_category, video_rating, video_description, 
    video_retrieved_date

(b) "video_tags.txt”
    This file contains a list of video tags for each video.

    Format: YoutubeID;\t tag1; tag2; ...; tagN

(c) "testset.txt"
    This file contains YouTube IDs of the videos used for evaluatation. We 
    provide these videos on:

    https://data.vision.ee.ethz.ch/cvl/video2gif/YouTubeID.mp4

(d) "./v2g_evaluation/"
    This directory contains evaluation code for nMSD and Average Precision used 
    in our CVPR 2016 paper. It requires Python packages 'numpy', 'scikit-learn' 
    and 'pandas' which can be installed using pip.

(e) "example.py"
    This script shows how to evaluate the predictions of a model.

(f) "setup.py"
    Script to install the evaluation package. Run `python setup.py install --user`

Last edit: August 10, 2017