Add a VAME view

[h-mayorquin]

This adds a viewer for VAME motif segmentation. When it finds a VAME object, the view reads the per-frame ndx-vame MotifSeries and run-length-encodes the motif sequence in the browser (no change to the file) into the in-memory equivalent of an EthogramBouts table: each run of a single motif becomes one (start_time, stop_time, label) row, i.e. the same dynamic-table / TimeIntervals shape a stored bouts table has. It then renders that as a montage over a session ethogram strip: pick a motif in the sidebar (or click its lane) and the montage plays several example clips of it side by side, looping in sync, cropped from the linked video with the pose keypoints and skeleton overlaid; when a file has pose but no video the clips play pose-only on a black canvas.

Rather than building this only for VAME, the montage-and-strip renderer is tool-agnostic: it consumes labeled behavioral bouts (the published ndx-ethogram EthogramBouts type, one row per bout with a single text label), and VAME is a thin adapter that produces those bouts from a MotifSeries. Because the renderer makes no VAME-specific assumptions, the same view serves any segmentation or classification tool that emits bouts, which is what lets it work directly on the public DANDI 001878 ("Standardized behavioral segmentation datasets") with no extra code. That dandiset stores each tool's output as EthogramBouts, including a VAME table, so the VAME motif output and several other tools all render through the one view:

• VAME motifs (public, 001878)
• keypoint-MoSeq syllables
• B-SOiD clusters
• CalMS21 named behaviors (video + pose)

[github-actions]

Preview deployment ready!

https://create-vame-view.neurosift.pages.dev

_{Built from e4523bf}