This is a meta issue that tracks additional segmenting algorithms as discussed in #40. Not all of these will be added, this issue is just a place to throw basically all algorithms we could possibly add.

Notice here I am grouping them by whether they produce Segments or a set of Boxes as discussed in #40, #49, #50, #51

segments

• whatever evsonganaly uses -- is there a name? amplitude_threshold
• ava (added in ENH: Add segment module with ava algorithm, fixes #40 #46 #53 to fix ENH: Add implementation of AVA method for segmenting mouse USVs #46 )
• DOC: Add vignette with segmenting algorithm from Ghaffari Devos 2023 + IR metrics #56
• usvseg: ENH: Add Python implementation of usvseg to vocalpy.signal.segment #43
• ENH: Add MUPET segmenting, thresholds on energy of gammatone spectrogram #116 to be able to re-use GoffiNet 🙂 dataset
• segment: amplitude variance from Alam Zia Roberts 2024 #168
• vocalseg -- "dynamic thresholding"
• endpoint_detection, AKA Rabiner-Sambur algorithm
• Du Troyer 2006 zebra finch algorithm -- gets cited a lot, I don't find any public implementations: https://www.sciencedirect.com/science/article/abs/pii/S0925231205004388?fr=RR-2&ref=pdf_download&rr=8095fa99e9b95848
• Harma

boxes

• Lasseck -- closest implementation I've found is https://github.com/NickleDave/BirdSongIdentification, we shouldn't copy directly but most of the steps are there.
• vocalseg -- the other one
• "ax" (acoustic segmenter): used with mouse USVs, https://github.com/JaneliaSciComp/Ax/blob/master/ax1.py
• method built into scikit-maad, find_rois, I think gives back boxes: https://scikit-maad.github.io/generated/maad.rois.find_rois_cwt.html#maad.rois.find_rois_cwt, even though it's described as "Temporal (1D)" (here). There's no single function under "Spectro-temporal (2D)" that gives back a similar list of ROIs along, AFAICT