Description

Related AliceVision pull request: alicevision/AliceVision#1457

This PR sets a dynamic size for the KeyframeSelection node, thus allowing the nodes that follow to be chunked if applicable.

As its number of outputs can rarely be precisely known before it is actually computed, we set the size with what we know to be the maximum number of keyframes.

There are 3 cases to determine what the maximum number of keyframes will be:

1. The number of input frames is known (the input is not a video) and the smart selection is enabled: the size is the minimum between the number of input frames and the maximum number of output frames for the smart selection.
2. The number of input frames is known and the regular selection is enabled:
   a. If the maximum number of output frames for the regular selection is set, then the size is the minimum between the number of input frames and the maximum number of output frames for the regular selection;
   b. If the maximum number of output frames for the regular selection is not set, we fall back to 1).
3. The number of input frames is not known (the input is a video): the size is the maximum number of output frames from the active selection mode.