How we are building a distributed multi-camera real-time sports tracking system using GStreamer and Rust

Display subtitles automatically

None
In the preferred language only when the video is using another
Always in the preferred language
Always in the original language of the video

Preferred language:

Qualities

UbiCast player x

Keyboard shortcuts

Key

Action

K or space

Play / Pause

Mute / Unmute

Select next subtitles

Select next audio track

Show slide in full page or toggle automatic source change

left arrow

Seek 5s backward

right arrow

Seek 5s forward

shift + left arrow or J

Seek 10s backward

shift + right arrow or L

Seek 10s forward

control + left arrow

Seek 60s backward

control + right arrow

Seek 60s forward

shift + down arrow

Decrease volume

shift + up arrow

Increase volume

shift + comma

Decrease playback rate

shift + dot or shift + semicolon

Increase playback rate

end

Seek to end

beginning

Seek to beginning

Loading Click here to add:

Information on this media

Number of views:

183 (this month: 25)

Creation date:

Sept. 26, 2023

Speakers:

Robin Gustavsson | Daniel Pendse

License:

CC BY-SA 3.0

Description

At Spiideo we offer automated sports video solutions to our customers for recording, analysis and broadcasting. We use a multi-camera setup to create a stitched panoramic video with an AI assisted cameraman. This talk will showcase how we moved from a segment based system with a glass-to-glass latency of almost two minutes to a frame-based system with a latency of around three seconds. In those three seconds we need to perform stitching of multiple 4K streams, detect objects in each camera stream and predict where to aim the virtual camera to follow the action on the pitch. And we do all of this across multiple instances in the cloud. This was Spiideo’s first real use of GStreamer and we will talk about what we struggled with, what helped us (a lot) and what we still do not really (really) understand.