DESED and DataSED precomputed caches for Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification

{"This":[0],"record":[1],"provides":[2],"the":[3,12,49,75,81,96,99,106,110],"precomputed":[4,53],"DESED":[5,101],"and":[6,46,52,59,67,102],"DataSED":[7,103],"dataset":[8],"caches":[9,29],"used":[10],"in":[11,15,109],"experiments":[13,79],"reported":[14],"\\"Cover":[16],"First,":[17],"Disagree":[18],"Softly:":[19],"Rethinking":[20],"Mismatch-First":[21],"Active":[22],"Learning":[23],"for":[24,48],"Frame-Level":[25],"Audio":[26],"Classification.\\"":[27],"The":[28],"contain":[30],"frame-level":[31],"audio":[32,86,90],"embeddings":[33,45],"extracted":[34,100],"using":[35,80],"PANNs":[36],"Cnn14_DecisionLevelMax,":[37],"corresponding":[38],"multi-label":[39],"annotations,":[40],"valid":[41],"frame":[42],"counts,":[43],"segment-level":[44],"labels":[47],"training":[50],"pool,":[51],"Euclidean":[54],"distance":[55],"matrices.":[56],"Training,":[57],"validation,":[58],"test":[60],"splits":[61],"are":[62],"included,":[63],"together":[64],"with":[65],"class":[66],"segment":[68],"metadata.":[69],"These":[70],"files":[71],"support":[72],"reproduction":[73],"of":[74],"paper's":[76],"active":[77],"learning":[78],"accompanying":[82],"code,":[83],"without":[84],"repeating":[85],"feature":[87],"extraction.":[88],"Raw":[89],"is":[91],"not":[92],"included.":[93],"To":[94],"use":[95],"caches,":[97],"place":[98],"directories":[104],"under":[105],"CACHE_ROOT":[107],"configured":[108],"code":[111],"repository.":[112]}

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-09-18
DOI
https://doi.org/10.5281/zenodo.22825177
Primary Topic
Music and Audio Processing
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
preprint

DESED and DataSED precomputed caches for Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification

Shiqi Zhang, Tuomas Virtanen
arXiv (Cornell University)
Music and Audio Processing
preprint

DESED and DataSED precomputed caches for Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification

Shiqi Zhang, Tuomas Virtanen
preprint en

Abstract

This record provides the precomputed DESED and DataSED dataset caches used in the experiments reported in "Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification." The caches contain frame-level audio embeddings extracted using PANNs Cnn14_DecisionLevelMax, corresponding multi-label annotations, valid frame counts, segment-level embeddings and labels for the training pool, and precomputed Euclidean distance matrices. Training, validation, and test splits are included, together with class and segment metadata. These files support reproduction of the paper's active learning experiments using the accompanying code, without repeating audio feature extraction. Raw audio is not included. To use the caches, place the extracted DESED and DataSED directories under the CACHE_ROOT configured in the code repository.

arXiv (Cornell University)
Music and Audio Processing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

DESED and DataSED precomputed caches for Cover First, Disagree Softly: Rethinking Mismatch-First Active Learning for Frame-Level Audio Classification — Shiqi Zhang, Tuomas Virtanen · arXiv (Cornell University) (2026) | TGRS Research Map | TGRS