Trial Set
GSC Short
Uses the Speech Commands testing split to build a short-duration SV protocol.
- Source dataset
- Google Speech Commands
- Trials
- 220000
- Protocol source
- Self-constructed protocol
- Source split
- Official testing split
Source Dataset
About Google Speech Commands
Speech Commands is a collection of short English recordings made for limited-vocabulary keyword recognition, not speaker verification. The commonly used v0.02 release contains 35 command words and about 106,000 one-second utterances. OpenSVBench uses speaker labels from the official testing split to build its own short-utterance speaker-verification trials.
- Language
- English
- Clip length
- About one second per utterance
- Vocabulary
- 35 command words in the v0.02 release
- Original task
- Small-vocabulary keyword recognition
Direct Ranking
Systems on this trial set
23 reviewed systems currently have valid results.
| Rank | System | Submitted by | Trial-Set EER ↓ | Trial-Set minDCF ↓ |
|---|---|---|---|---|
| #1 |
Agora Voiceprint (ResNet293)
agora_vp
|
agora_vp | 2.5250 | 0.147370 |
| #2 |
Dual-LoRA W2V-BERT2-Adapter-MFA (TidyVoice)
duallora-w2vbert2-tidyvoice-full-core
|
shangguanqituan | 4.2715 | 0.285070 |
| #3 |
iic/eres2netv2-zh
eresv2_zh
|
Public checkpoint | 5.0750 | 0.303175 |
| #4 |
wespeaker/w2vbert2
vb2-w2vbert
|
Public checkpoint | 6.0250 | 0.388025 |
| #5 |
iic/campplus
campplus
|
Public checkpoint | 7.0055 | 0.425695 |
| #6 |
W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2)
Wav2Vec2-BERT
|
pengjy | 7.0645 | 0.517445 |
| #7 |
palabraai/redimnet2-b6-vb2+vox2
redimnet2-b6-vb2vox2-lm-full-core
|
Public checkpoint | 8.1985 | 0.487250 |
| #8 |
wespeaker/res34-cnceleb
wespeaker_r34
|
Public checkpoint | 8.3650 | 0.496165 |
| #9 |
AntSpeaker/MECT-B2
antspeaker-mect-b2-vb2-full-core
|
Public checkpoint | 8.8100 | 0.492470 |
| #10 |
wespeaker/res152-voxceleb
wespeaker_resnet152
|
Public checkpoint | 10.8860 | 0.601260 |
| #11 |
wespeaker/samresnet100
vb2-sam100
|
Public checkpoint | 11.2450 | 0.611560 |
| #12 |
wespeaker/res293-voxceleb
wespeaker_resnet293
|
Public checkpoint | 11.4500 | 0.608915 |
| #13 |
palabraai/redimnet2-b6
redimnet2
|
Public checkpoint | 12.3950 | 0.640440 |
| #14 |
wespeaker/ecapa512-voxceleb
wespeaker_ecapa512
|
Public checkpoint | 12.8770 | 0.644885 |
| #15 |
wespeaker/res34-voxceleb
wespeaker_resnet34_voxceleb
|
Public checkpoint | 13.2190 | 0.705995 |
| #16 |
iic/eres2net-large-3dspeaker
eres2net_large_3dspeaker
|
Public checkpoint | 13.3500 | 0.658090 |
| #17 |
iic/eres2net-en
eres_en
|
Public checkpoint | 13.8900 | 0.693170 |
| #18 |
wespeaker/campplus-voxceleb
campplus_wespeaker
|
Public checkpoint | 16.4345 | 0.764210 |
| #19 |
speechbrain/ecapa
ecapa
|
Public checkpoint | 17.0600 | 0.777175 |
| #20 |
wespeaker/ecapa1024-voxceleb
wespeaker_ecapa1024
|
Public checkpoint | 17.4110 | 0.776190 |
| #21 |
speechbrain/xvector
xvector
|
Public checkpoint | 17.5650 | 0.848580 |
| #22 |
microsoft/unispeech-sat-base-plus-sv
unispeech_sat_base_plus
|
Public checkpoint | 24.4600 | 0.966790 |
| #23 |
microsoft/wavlm-base-plus-sv
wavlm_base
|
Public checkpoint | 25.7750 | 0.967470 |