iVID-Datasets for AVC (H.264) and HEVC (H.265)

DASH Datasets for AVC (H.264) and HEVC (H.265)

Datasets for AVC (H.264) and HEVC (H.265) Evaluation of Dynamic Adaptive Streaming over HTTP (DASH)


We kindly ask that should you mention any of our datasets, or use our code, in your publication, that you would reference the following paper:

Jason J. Quinlan, Ahmed H. Zahran, Cormac J. Sreenan, "Datasets for AVC (H.264) and HEVC (H.265) Evaluation of Dynamic Adaptive Streaming over HTTP (DASH)". In Proceedings of the 7th ACM Multimedia Systems Conference 2016, Klagenfurt am Wörthersee, Austria. May 10-13, 2016. Further details on our datasets are available in the conference paper.

The following provides an overview of our three datasets, plus additional information for stream evaluation and clip content:

  1. DATASET 1: Testbed Dataset - the testbed evaluation dataset offers a means of analysing the delivery of content over a physical network and associated adaptation mechanisms at the client. This is non copyright DASH content, which can be streamed to a compatible player and the following table provides links to the open-source MPD files for each of these three clips over multiple segment durations and across H.264 and H.265. 

    2-second segment 4-second segment 6-second segment 8-second segment 10-second segment
    bbb-264 bbb-264 bbb-264 bbb-264 bbb-264
    ed-264 ed-264 ed-264 ed-264 ed-264
    sstb-264 sstb-264 sstb-264 sstb-264 sstb-264
    bbb-265 bbb-265 bbb-265 bbb-265 bbb-265
    ed-265 ed-265 ed-265 ed-265 ed-265
    sstb-265 sstb-265 sstb-265 sstb-265 sstb-265
  2. DATASET 2: Trace-based Dataset - the trace-based simulation dataset provides a means of stream evaluation in frameworks such as NS-2 and NS-3. The Segment Distribution files for all twenty three video clips can be found here.

    To provide a means of quantifying the QOE for each streamed clip using the Trace-based Dataset, we also provide PSNR results, for each of the clips, for each segment per representation rate, which can be found here. By determining which representation rate per segment is downloaded, the relative PSNR values can be summed to determine an average PSNR value per clip.

  3. DATASET 3: Header Dataset - The Header or Dummy dataset, which can be found here, is composed of the header information from each of our two hundred and thirty evaluated clips (twenty three clips * two encoders * five segment durations). As we are not making the content available and only the decoding parameters in the header files, we can release the header information for both our testbed evaluation and trace-based datasets. Per clip, we offer the respective MPD file, the header information for the MP4 file as well as the header information per segment. The instructions for building the complete Dataset are clearly detailed within the zip file, see the READ_ME.txt file.

  4. In our Evaluation we utilise a media player called MP4Client to stream the "Testbed Dataset" and "Header Dataset". Instruction for building GPAC, a media framework containing MP4Client, and the use of same can be found here. Instructions for running GPAC and extracting the log files are clearly detailed within the zip file, see the READ_ME.txt file.

  5. We also provide a VirtualBox VM of a fresh install of Ubuntu 14.04, complete with all required dependencies for H.264 and H.265 decoding and streaming using GPAC: Ubuntu VM. The user name and password for the VM are both "ubuntu". We have also installed the relevant "Guest Additions" to the VM.

  6. The following table provides details on each of the video clips in the dummy dataset:

    Clip Number Genre Available Content Scene Description Clip Duration
    Clip 1 Sci-Fi Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes in a desert scenario 16 minutes
    Clip 2 TV Show Copyrighted - Trace-based Dataset, PSNR and Header Dataset Mixture of slow moving and static scenes, travel documentary style 16 minutes
    Clip 3 Sci-Fi Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, but with highly variable background 16 minutes
    Clip 4 Animated Open Source Content - Big Buck Bunny - all Content Available Slow moving scenes 10 minutes
    Clip 5 Animated Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, with a mixture of highly variable to static background 16 minutes
    Clip 6 Thriller Copyrighted - Trace-based Dataset, PSNR and Header Dataset Highly variable dynamic scenes with fast moving central characters 16 minutes
    Clip 7 Animated Open Source Content - Elephant Dreams Fast Moving Scenes 10 minutes
    Clip 8 Comedy Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 9 Fantasy Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, typically with static backgrounds 16 minutes
    Clip 10 Sci-Fi Copyrighted - Trace-based Dataset, PSNR and Header Dataset Highly variable dynamic scenes with fast moving central characters 16 minutes
    Clip 11 Adventure Copyrighted - Trace-based Dataset, PSNR and Header Dataset Highly variable dynamic scenes with fast moving central characters 16 minutes
    Clip 12 TV Show Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 13 Documentary Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, typically with static backgrounds 16 minutes
    Clip 14 TV Show Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 15 Animated Open Source Content - Sita Sings the Blues Slow moving scenes 16 minutes
    Clip 16 Sci-Fi Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, mixture of static and variable backgrounds 16 minutes
    Clip 17 Adventure Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 18 Thriller Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, typically slow moving background 16 minutes
    Clip 19 Fantasy Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 20 Animated Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow moving scenes, typically with static backgrounds 16 minutes
    Clip 21 Animated Copyrighted - Trace-based Dataset, PSNR and Header Dataset Highly variable dynamic scenes with fast moving central characters 16 minutes
    Clip 22 Documentary Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, typically with static backgrounds 16 minutes
    Clip 23 Comedy Copyrighted - Trace-based Dataset, PSNR and Header Dataset Slow and fast moving scenes, typically with static backgrounds 16 minutes

Enjoy

This work has emanated from research conducted with the financial support of Science Foundation Ireland (SFI) under Grant Number 13/IA/1892.

Mobile and Internet Systems Laboratory

Department of Computer Science, Western Gateway Building, University College Cork, Western Road, Cork, Ireland.

Top