Machine Learning - Part 8 - Serialization, Caching & Archiving - Flame 2021.2

The Flame Learning Channel

Indhold leveret af Grant Kay. Alt podcastindhold inklusive episoder, grafik og podcastbeskrivelser uploades og leveres direkte af Grant Kay eller deres podcastplatformspartner. Hvis du mener, at nogen bruger dit ophavsretligt beskyttede værk uden din tilladelse, kan du følge processen beskrevet her https://da.player.fm/legal.

4y ago 6:29

M4V•Episode hjem

In part 8 of the Machine Learning series, we examine a series of performance enhancements for Machine Learning and internally generated Motion Vectors in the Timeline in the Flame 2021.2 Update.

When using Machine Learning Models and generated Motion Vectors Maps, Flame will now cache the frames to disk to enhanced performance. This removes the need to recalculate the data analysis each time you visit the TimelineFX.

Since this type of caching is now part of the project, it can now be managed as well as archived with the Flame project!

There is also a new notification when Machine Learning models are initialising to let you know what Flame is doing.

And finally, as a new CentOS only feature, the machine learning models are serialised or written to your system disk in order to load faster when you need them. To be clear, this affects the loading times for the machine learning algorithm and not the computations when analysing the footage with a machine learning model.

447 episoder