System Detail
wespeaker/ecapa1024-voxceleb
wespeaker_ecapa1024
Submitted by: Public checkpoint
- Global position
- #16 / 23
- Scenario-macro EER
- 14.0835
- Scenario-macro minDCF
- 0.534111
- Trial-set coverage
- 26 / 26
- Scenario top-three
- 0
Publicly ranked
This result is included in the reviewed public ranking.
Model Provenance
Source and architecture
- Displayed name:
wespeaker/ecapa1024-voxceleb - Submitted by: Public checkpoint
- Metadata: OpenSVBench model record
- Source: WeSpeaker
- Architecture: ECAPA-TDNN
- Model family: ECAPA
- Submitted alias:
wespeaker-voxceleb-ecapa-tdnn1024-LM - Created at:
2026-05-29T14:49:19.687079+00:00
Training And Links
Training data and references
- Training data: VoxCeleb2 dev, 5,994 speakers
- Training setup: ECAPA-TDNN GLOB c1024 with ASTP pooling, 192-d embedding, ArcMargin, 150-epoch training, speed perturbation, and MUSAN/RIRS augmentation.
- Submission mode:
full-core - Model source: Open link
- Paper: ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification
Higher-Ranked Scenarios
Scenario ranks above the global position
Show 3 supporting trial sets
- CHiME-6 Overlap: #13 on its trial-set ranking.
- VoxCeleb1-O: #14 on its trial-set ranking.
- ESD: #14 on its trial-set ranking.
Lower-Ranked Scenarios
Scenario ranks below the global position
Show 3 supporting trial sets
- CN-Celeb Genre: #20 on its trial-set ranking.
- CN-Celeb Short: #20 on its trial-set ranking.
- GSC Short: #20 on its trial-set ranking.
Scenario Rankings
Full ranking across scenarios
This table shows where the model sits on each scenario, using source-balanced scenario scores and the linked trial sets as evidence.
| Scenario | Rank | Lens | Evidence | Score |
|---|---|---|---|---|
|
Aging Robustness
How stable identity representations remain across age-derived and longitudinal recording gaps.
|
#14/23
Needs work
|
Speaker aging and time gaps
2.2406 EER from leader
|
4.0323
0.181572 minDCF
|
|
|
Speaking-Style Robustness
Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.
|
#15/23
Needs work
|
Speaking style shift
4.6525 EER from leader
|
6.5691
0.361235 minDCF
|
|
|
Noise / Reverb Robustness
How reliably the model preserves identity when room acoustics, distractor noise, and microphone placement deviate from an easier in-corpus reference condition.
|
#15/23
Needs work
|
Noise and reverberation
4.3460 EER from leader
|
6.6060
0.214720 minDCF
|
|
|
Overlap Robustness
How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.
|
#15/23
Needs work
|
Overlapping speakers
6.5318 EER from leader
|
14.6411
0.618143 minDCF
|
|
|
Cross-Lingual Robustness
How well speaker identity survives enrollment-test language mismatch.
|
#17/23
Needs work
|
Language mismatch
5.3271 EER from leader
|
6.9847
0.436029 minDCF
|
|
|
Distance Robustness
How much performance changes across meeting and domestic distance mismatch.
|
#17/23
Needs work
|
Distance mismatch
10.1872 EER from leader
|
15.4440
0.640785 minDCF
|
|
|
Accent / Dialect Robustness
How reliably the model tracks identity across accent and dialect mismatch.
|
#17/23
Needs work
|
Accent and dialect variation
7.4947 EER from leader
|
24.3163
0.734606 minDCF
|
|
|
In-The-Wild Robustness
How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.
|
#18/23
Needs work
|
Open-domain media speech
4.7987 EER from leader
|
8.5204
0.308148 minDCF
|
|
|
Channel / Device Robustness
How resilient the model is to device and channel mismatch.
|
#18/23
Needs work
|
Device and channel variation
14.4543 EER from leader
|
18.4567
0.824736 minDCF
|
|
|
Short-Duration Robustness
How much performance holds up when speech evidence is limited by duration.
|
#20/23
Needs work
|
Short-duration speech
9.2104 EER from leader
|
18.2450
0.644928 minDCF
|
|
|
Genre-Shift Robustness
Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.
|
#20/23
Needs work
|
Source-genre variation
16.8300 EER from leader
|
31.1025
0.910323 minDCF
|
Variant Compare
Compare ECAPA variants
Expand this section to compare models in the same model family by architecture, source, training data, training setup, and leaderboard result.
Show
Compare ECAPA variants
Expand this section to compare models in the same model family by architecture, source, training data, training setup, and leaderboard result.
| Variant | Submitted by | Source | Training Data | Training Setup | Global Rank | Macro EER | Macro minDCF | Open |
|---|---|---|---|---|---|---|---|---|
|
speechbrain/ecapa
ECAPA-TDNN
|
Public checkpoint | SpeechBrain | VoxCeleb1 + VoxCeleb2 training data | SpeechBrain ECAPA-TDNN release with attentive statistical pooling and Additive Margin Softmax loss. |
#15
ranked
|
13.5461 | 0.543660 | Open |
|
wespeaker/ecapa1024-voxceleb
ECAPA-TDNN
Current
|
Public checkpoint | WeSpeaker | VoxCeleb2 dev, 5,994 speakers | ECAPA-TDNN GLOB c1024 with ASTP pooling, 192-d embedding, ArcMargin, 150-epoch training, speed perturbation, and MUSAN/RIRS augmentation. |
#16
ranked
|
14.0835 | 0.534111 | Open |
|
wespeaker/ecapa512-voxceleb
ECAPA-TDNN
|
Public checkpoint | WeSpeaker | VoxCeleb2 dev, 5,994 speakers | ECAPA-TDNN GLOB c512 with ASTP pooling, 192-d embedding, ArcMargin, large-margin fine-tuning, speed perturbation, and MUSAN/RIRS augmentation. |
#18
ranked
|
14.3120 | 0.541141 | Open |
Trial-Set Results
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
Show
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
| Trial Set | Rank | Scenarios | Trials | Score |
|---|---|---|---|---|
|
TidyVoiceX2-ASV
TidyVoiceX2-ASV
|
#17/23
|
Cross-lingual
|
200000 |
6.9847
0.436029 minDCF
|
|
HI-MIA
HI-MIA
|
#19/23
|
Short-duration
|
660000 |
8.6250
0.381742 minDCF
|
|
GSC Short
Google Speech Commands
|
#20/23
|
Short-duration
|
220000 |
17.4110
0.776190 minDCF
|
|
GLOBE
GLOBE
|
#16/23
|
Accent/dialect
|
56848 |
25.7740
0.541118 minDCF
|
|
CN-Celeb
CN-Celeb
|
#19/23
|
In-the-wild
|
3484292 |
15.8209
0.536618 minDCF
|
|
CN-Celeb Genre
CN-Celeb
|
#20/23
|
Genre shift
|
440000 |
31.1025
0.910323 minDCF
|
|
CN-Celeb Short
CN-Celeb
|
#20/23
|
Short-duration
|
546964 |
30.3869
0.901376 minDCF
|
|
VoxCeleb1-O
VoxCeleb1
|
#14/23
|
In-the-wild
|
37611 |
0.8188
0.065623 minDCF
|
|
VoxCeleb1-E
VoxCeleb1
|
#15/23
|
In-the-wild
|
579818 |
0.9899
0.062012 minDCF
|
|
VoxCeleb1-H
VoxCeleb1
|
#15/23
|
In-the-wild
|
550894 |
1.8513
0.111398 minDCF
|
|
VoxCeleb Short
VoxCeleb1
|
#16/23
|
Short-duration
|
394724 |
16.5570
0.520405 minDCF
|
|
3D-Speaker Device
3D-Speaker
|
#18/23
|
Channel/device
|
180000 |
23.8800
0.940667 minDCF
|
|
3D-Speaker Distance
3D-Speaker
|
#18/23
|
Distance
|
175163 |
23.6657
0.920046 minDCF
|
|
3D-Speaker Dialect
3D-Speaker
|
#17/23
|
Accent/dialect
|
180000 |
22.8587
0.928093 minDCF
|
|
FFSVC 2022 Cross-Channel
FFSVC 2022
|
#20/23
|
Channel/device
|
72000 |
13.0333
0.708806 minDCF
|
|
FFSVC 2022 Cross-Domain
FFSVC 2022
|
#20/23
|
Distance
|
66546 |
13.1585
0.719381 minDCF
|
|
Whisper40 Whisper
Whisper40
|
#17/23
|
Speaking style
|
17600 |
14.8000
0.736062 minDCF
|
|
Lombard Grid Lombard
Lombard Grid
|
#15/23
|
Speaking style
|
29524 |
0.3055
0.017288 minDCF
|
|
VOiCES Noise/Reverb
VOiCES
|
#15/23
|
Noise/reverb
|
55000 |
6.6060
0.214720 minDCF
|
|
CHiME-6 Domestic Far-Field
CHiME-6
|
#14/23
|
Distance
|
36487 |
13.6569
0.577058 minDCF
|
|
CHiME-6 Overlap
CHiME-6
|
#13/23
|
Overlap
|
39600 |
17.8889
0.675139 minDCF
|
|
ESD
ESD
|
#14/23
|
Speaking style
|
437408 |
4.6018
0.330355 minDCF
|
|
AliMeeting Near/Far
AliMeeting
|
#19/23
|
Distance
|
220000 |
11.2950
0.346655 minDCF
|
|
AliMeeting Overlap
AliMeeting
|
#18/23
|
Overlap
|
165000 |
11.3933
0.561147 minDCF
|
|
VoxKnesset
VoxKnesset
|
#14/23
|
Aging
|
158312 |
5.6932
0.252750 minDCF
|
|
VoxPopuli Aging
VoxPopuli
|
#15/23
|
Aging
|
146575 |
2.3715
0.110394 minDCF
|