Brightcove Streaming Datasets

This project contains several examples of real-world streaming playback statistics provided for academic research purposes. The description of this data set and possible examples of its use are explained in papers:

  • T. Teixeira, B. Zhang, Y. Reznik, "Adaptive Streaming Playback Statistics Dataset," Proc. 12th ACM Multimedia Systems Conference (MMSys '21), Istanbul, Turkey. September 28-October 1, 2021. DOI 10.1145/3458305.3478444

  • Y. Reznik, K. Lillevold, A. Jagannath, and X. Li, "Towards Understanding of the Behavior of Web Streaming," Proc. The Picture Coding Symposium (PCS'21), Bristol, UK, June 29 - July 2, 2021.

This material is shared under Apache License Version 2.0, with the full text of the license provided in file LICENSE.

Explanation of parameters/data fields provided in the data set:

Files are organized in directories, one for each event. Files inside those directories are organized with the following naming convention:

eventX_playback_statistics_Y.csv

where X is the event number or identification and Y is the file number breakdown for each event to accommodate GitHub's limitation of 100 MB file size.

Each file contains the following columns:

Category Parameter Description
session session Randomly generated number associated with each player session
seq Sequential number of an event within a session
client device_type Device type. Possible values: “desktop”, “mobile”, “tablet”, “tv”, “other”
device_os OS type. Possible values: “windows”, “osx”, “linux”, “android”, “ios”, ”webos”, ”other”
browser Browser type. Possible values: “chrome”, “firefox”, “safari”, “edge”, “ie”, “opera”, “other”
player Player type. Possible values: “app” – dedicated application, “web” – JS / browser-based player
player_width Player window width [pixels]
player_height Player window height [pixels]
rendition rendition_indicated_bps Rendition bitrate [bps]. Sum of audio and video bitrates.
rendition_width Video width as encoded [pixels]
rendition_height Video height as encoded [pixels]
rendition_framerate Video framerate [fps]
video_codec Video codec type. Possible values: “h264”, “hevc”, “av1”
video_codec_profile Video codec profile. Possible values: “baseline”, “main”, “high”
format Streaming format. Possible values: “hls_v3”, “hls_v7”, “dash”
segment_duration Segment duration [seconds]
playback video_seconds_viewed Seconds of media content played in the period between the last two player events
forward_buffer_seconds The number of seconds of media content buffered but not yet played
rebuffering_seconds The total number of seconds the player was “buffering” in the period between the last two player events
rebuffering_count The number of times the player was “buffering” in the period between the last two player events
media_bytes_transferred The total number of bytes transferred since the start of the session
network measured_bps Network bandwidth [bps] estimated based on size and delivery time of the last segment downloaded

Citing our Work

If you found the datasets useful and used in your research, we kindly ask that you would reference the following paper:

T. Teixeira, B. Zhang, Y. Reznik, "Adaptive Streaming Playback Statistics Dataset," Proc. 12th ACM Multimedia Systems Conference (MMSys '21), September 28-October 1, 2021, Istanbul, Turkey

Contact the Authors

Thiago Teixeira (tteixeira at brightcove dot com)
Bo Zhang
Yuriy Reznik