1427bd8e8a5b — Chris Cannam default tip 29 days ago
Updated API
784b5dd1099f — Chris Cannam 2 months ago
Add support for step factors
a87bdf73e7a8 — Chris Cannam 2 months ago
Important fix to calculateAdvance - don't advance both if row is just one short, as this can happen spuriously
5ef0eb09e8f6 — Chris Cannam 2 months ago
I think this will have to do
d1d92ad00256 — Chris Cannam 2 months ago
Types were not quite correct - confusion between mono and poly vectors. Could usefully improve this. Add a full-size file-backed Real variant (which is what exposed this)
Merge from branch filebacked
Switch entire state when complete, so that all working data can be garbage-collected; add tempDir
Use "live" vectors when still calculating forward face; store thereafter. Much faster
Handle deleting costContainer properly
Concrete file-backed versions
Sketch out use of a file-backed vector
Make it possible to construct constrained DTWs with on-demand feature vectors
8947e8e95a89 — Chris Cannam 2 years ago
Add silence threshold; print costs from constrained-only runs too
04db0be81fa3 — Chris Cannam 2 years ago
Include cost output from "both" mode as well
a107d4c0f6be — Chris Cannam 2 years ago
Colour images, and a bit more info output & alignment tweaking
26ee84347dec — Chris Cannam 2 years ago
Further cost reporting; normalise in example
0964cd64a25d — Chris Cannam 2 years ago
Add function to retrieve total cost
76b247b09ff1 — Chris Cannam 2 years ago
Report stats
551cc9c16c0d — Chris Cannam 2 years ago
Simplify, & restore image saving with MLton compatibility
0fbabb4be47b — Chris Cannam 2 years ago
Backed out changeset 67a761f7d2b1. Turns out there is nothing to gain at all, because we need the lengths of both feature vectors at the outset - even the constrained DTW isn't (yet?) written in a way that avoids this - and we don't know those without completely filling both
Next