List your company
ResearchFormerly WizWand

Speaker Diarization on AliMeeting (test)

0.0559DERTSVAD
0.0498560.0906530.131450.172247Mar 18, 2022Dec 2, 2022Aug 19, 2023May 5, 2024Jan 19, 2025Oct 6, 2025Jun 23, 2026
Updated 4mo ago

Evaluation results

MethodLinks
TSVADTraining set=sim.+real, Size=21.16M, FLOPS=60.80M2022.030.0559
SENDTraining set=sim.+real, Size=18.42M, FLOPS=36.73M2022.030.058
SENDTraining set=sim., Size=18.42M, FLOPS=36.73M2022.030.0811
TSVADTraining set=sim., Size=21.16M, FLOPS=60.80M2022.030.087
SA-S2SND (E7 in Table 2)Evaluation Scenario=Offline, Model Size=Medium(45.96M+0.86M), Channels=1, Training Data=Compound dataset, Usage of DOA=true2025.100.104
MS-SphereVBx-PFPipeline=EEND-VC2026.060.106
SphereVBxPipeline=EEND-VC2026.060.107
MS-SphereVBxPipeline=EEND-VC2026.060.107
WavLM-LargeEvaluation Scenario=Offline2025.100.108
BaselinePipeline=EEND-VC2026.060.108
SOTA 02/2026Pipeline=EEND-VC2026.060.108
S2SND (E6 in Table 2)Evaluation Scenario=Offline, Model Size=Medium(45.96M), Channels=1, Training Data=Compound dataset2025.100.1133
EEND-TA + FTEvaluation Scenario=Offline2025.100.1141
EEND-EDAEvaluation Scenario=Offline2025.100.123
EEND-M2FEvaluation Scenario=Offline2025.100.132
Pyannote.audioEvaluation Scenario=Offline2025.100.152
VBxTraining set=sim.+real2022.030.156
DiaperEvaluation Scenario=Offline2025.100.207