Re: [rmcat] Test video sequence sources

[Sergio Mena <[e-mail]>]

Thomas,

First of all, we would like to thank you for the pointer you provide.

We have indeed explored the sequences available at xiph.org site in the 
past.

Unfortunately the vidyo sequences (vidyo1, vidyo2, and vidyo3), do not 
fulfill 2 of our requirements for a sequnce that can be used for 
obtaning video traces. Those sequences are too short (they are around 10 
secs long... we need 1 minute as a minimum, ideally 2 mins); they also 
have 720p resolution, so the traces obtained couldn't be used for 
simulating encoded 1080p video.

Indeed, the long sequences we are using at the moment are also available 
at the page you point at (big_buck_bunny and elephants_dream). The 
problem with those sequences is that they are animations, so when we 
encoded them, we could see huge peaks at some frames sizes in the 
encoded traces (see the slides presented in IETF95 for further info).

[As a reminder, our requirements are: (1) 1 min minimum length, ideally 
2 min, (2) conferencing-style content, i.e., no scene cuts, (3) 1080p 
min resolution. We are conscious that these requirement at somewhat 
stringent (uncompressed video will amount to 10s of GB), so we are 
planning on generating our own sequence (didn't have the resources to do 
so, so far)]

Thanks,

Sergio


On 19/07/16 12:09, Thomas Daede wrote:
> Xiph.Org hosts a large selection of test video sequences, donated from
> various participants:
>
> https://media.xiph.org/video/derf/
>
> There are longer videoconferencing sequences here, such as the Vidyo
> ones (600 frames), as well as many other types of content.
>
> - Thomas
>
>