Shot detection limit or quota

[PlanetDevTeam]

Hello, while trying to execute a shot detection on a video that is 8 hours I got to understand that there's a processing limit of 3 hours. I would like to understand if this is a technical limit or if it's a quota, but since it's not listed in the quotas page I guess it's something "special"?

If it was possible to raise this limit to 8 hours or beyond it would greatly improve my job, is there anything that there could be done?

[Ismail (Cloud Platform Support)]

This is the 'Back time in seconds per minute' request quota. In order for you to increase your quota, you would need to submit a request through your Google Cloud Console by selecting Edit Quotas from the Quotas page. 

[PlanetDevTeam]

Thank you for your answer. In this page https://cloud.google.com/video-intelligence/quotas it says that "Backend time in seconds per minute" is "the amount of parallel processing time used to annotate video content", while I'd like to increase "Video API annotate video length". Are these configurations affected both by this same parameter? Is there a direct relationship?

[Jofre Riba Sanchez]

As explained in [1], "Video API annotate video length" is a limit (like "Video size"), while "Backend time in seconds per minute" is a quota.

The first sentence under "Request quotas" indicates that quotas are subject to change, which implies that the API limits are not subject to change.

This is the reason why quotas appear in the console's quota page, but limits do not: quotas can be changed for each user, while limits are part of the design of the API.

This means that, currently, it is not possible to execute shot detection in an 8h video. 

If you split the video and want to process the chunks it in parallel, then the "Backend time in seconds per minute" will come into play. If you split the 8h video in 3 chunks of 2.7h each, the default quota will be enough to process in parallel. If you split the video in 4 chunks of 2h each, the default quota will not be enough to process in parallel. You either have to queue one of the chunks, or request a quota increase for "Backend time in seconds per minute".

[1] https://cloud.google.com/video-intelligence/quotas