System Detail

wespeaker/ samresnet100

vb2-sam100

ranked review: approved WeSpeaker SAM-ResNet100

This page shows the global rank, scenario ranks, and trial-set results for this submission.

Back To Leaderboard Submit Another Run

Global Position

#3/18

Public ranking position among reviewed full-core submissions.

Scenario-Macro EER

10.5780

Scenario-Macro minDCF

0.390239

Coverage

26/26

1 scenario ranked #1 6 scenarios in top 3

Public Listing

Ranked Publicly

This full-core result is approved and included in the public ranking.

Model Provenance

Source and architecture

Displayed name: wespeaker/samresnet100
Source: WeSpeaker
Architecture: SAM-ResNet100
Submitted alias: voxblink2_samresnet100_ft
Created at: 2026-05-29T14:22:47.375445+00:00

Training And Links

Training data and references

Training data: VoxBlink2 (pretrain) + VoxCeleb2 (finetune)
Training setup: SimAM-ResNet100 ASP release fine-tuned on VoxCeleb2 with ArcMargin, speed perturbation, and MUSAN/RIRS augmentation.
Submission mode: full-core
Paper: Deep Residual Learning for Image Recognition

Open Source Checkpoint Open Paper

Higher-Ranked Scenarios

Scenario ranks above the global position

Aging #1 Noise/reverb #2 Cross-lingual #2

Show 3 supporting trial sets

Lombard Grid Lombard: #1 on its trial-set ranking.
VoxKnesset: #1 on its trial-set ranking.
VoxCeleb1-O: #2 on its trial-set ranking.

Lower-Ranked Scenarios

Scenario ranks below the global position

Accent/dialect #7 Channel/device #6 Genre shift #4

Show 3 supporting trial sets

3D-Speaker Distance: #8 on its trial-set ranking.
3D-Speaker Dialect: #7 on its trial-set ranking.
CN-Celeb Short: #6 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.	#1/18 Leader	Speaker aging and time gaps 0.0000 EER from leader	#1 2.5933 #3 1.5460	2.0696 0.068782 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.	#2/18 Top 3	Noise and reverberation 0.4000 EER from leader	#2 2.9800	2.9800 0.098660 minDCF
Cross-Lingual Robustness How well speaker identity survives enrollment-test language mismatch.	#2/18 Top 3	Language mismatch 0.2501 EER from leader	#2 4.7178	4.7178 0.362506 minDCF
In-The-Wild Robustness How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.	#2/18 Top 3	Open-domain media speech 0.0653 EER from leader	#5 9.9803 #2 0.2287 #3 0.4564 #2 0.8665	5.2487 0.170274 minDCF
Speaking-Style Robustness Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.	#3/18 Top 3	Speaking style shift 1.5147 EER from leader	#2 2.7503 #6 10.1250 #1 0.0373	4.3042 0.238095 minDCF
Overlap Robustness How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.	#3/18 Top 3	Overlapping speakers 0.9529 EER from leader	#5 8.4047 #2 13.7778	11.0912 0.502019 minDCF
Distance Robustness How much performance changes across meeting and domestic distance mismatch.	#4/18 Competitive	Distance mismatch 1.9041 EER from leader	#8 18.4835 #5 8.5605 #2 8.4112 #6 9.2164	11.1679 0.421520 minDCF
Short-Duration Robustness How much performance holds up when speech evidence is limited by duration.	#4/18 Competitive	Short-duration speech 0.9335 EER from leader	#6 5.9033 #6 11.2450 #6 25.4042 #3 11.2390	13.4479 0.497755 minDCF
Genre-Shift Robustness Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.	#4/18 Competitive	Source-genre variation 10.8475 EER from leader	#4 25.1200	25.1200 0.674765 minDCF
Channel / Device Robustness How resilient the model is to device and channel mismatch.	#6/18 Competitive	Device and channel variation 7.7067 EER from leader	#5 18.9367 #6 9.0972	14.0169 0.591790 minDCF
Accent / Dialect Robustness How reliably the model tracks identity across accent and dialect mismatch.	#7/18 Competitive	Accent and dialect variation 1.7588 EER from leader	#2 24.8065 #7 19.5813	22.1939 0.666463 minDCF

Trial-Set Results

Raw ranking by trial set

Expand this section to inspect the detailed trial-set rankings.

Show

Trial Set	Rank	Scenarios	Trials	Score
TidyVoiceX2-ASV TidyVoiceX2-ASV	#2/18	Cross-lingual	200000	4.7178 0.362506 minDCF
HI-MIA HI-MIA	#6/18	Short-duration	660000	5.9033 0.273133 minDCF
GSC Short Google Speech Commands	#6/18	Short-duration	220000	11.2450 0.611560 minDCF
GLOBE GLOBE	#2/18	Accent/dialect	56848	24.8065 0.531753 minDCF
CN-Celeb CN-Celeb	#5/18	In-the-wild	3484292	9.9803 0.311502 minDCF
CN-Celeb Genre CN-Celeb	#4/18	Genre shift	440000	25.1200 0.674765 minDCF
CN-Celeb Short CN-Celeb	#6/18	Short-duration	546964	25.4042 0.714774 minDCF
VoxCeleb1-O VoxCeleb1	#2/18	In-the-wild	37611	0.2287 0.011966 minDCF
VoxCeleb1-E VoxCeleb1	#3/18	In-the-wild	579818	0.4564 0.026098 minDCF
VoxCeleb1-H VoxCeleb1	#2/18	In-the-wild	550894	0.8665 0.049074 minDCF
VoxCeleb Short VoxCeleb1	#3/18	Short-duration	394724	11.2390 0.391553 minDCF
3D-Speaker Device 3D-Speaker	#5/18	Channel/device	180000	18.9367 0.711080 minDCF
3D-Speaker Distance 3D-Speaker	#8/18	Distance	175163	18.4835 0.669019 minDCF
3D-Speaker Dialect 3D-Speaker	#7/18	Accent/dialect	180000	19.5813 0.801173 minDCF
FFSVC 2022 Cross-Channel FFSVC 2022	#6/18	Channel/device	72000	9.0972 0.472500 minDCF
FFSVC 2022 Cross-Domain FFSVC 2022	#6/18	Distance	66546	9.2164 0.478007 minDCF
Whisper40 Whisper Whisper40	#6/18	Speaking style	17600	10.1250 0.485938 minDCF
Lombard Grid Lombard Lombard Grid	#1/18	Speaking style	29524	0.0373 0.004657 minDCF
VOiCES Noise/Reverb VOiCES	#2/18	Noise/reverb	55000	2.9800 0.098660 minDCF
CHiME-6 Domestic Far-Field CHiME-6	#2/18	Distance	36487	8.4112 0.319174 minDCF
CHiME-6 Overlap CHiME-6	#2/18	Overlap	39600	13.7778 0.612972 minDCF
ESD ESD	#2/18	Speaking style	437408	2.7503 0.223690 minDCF
AliMeeting Near/Far AliMeeting	#5/18	Distance	220000	8.5605 0.219880 minDCF
AliMeeting Overlap AliMeeting	#5/18	Overlap	165000	8.4047 0.391067 minDCF
VoxKnesset VoxKnesset	#1/18	Aging	158312	2.5933 0.101669 minDCF
VoxPopuli Aging VoxPopuli	#3/18	Aging	146575	1.5460 0.035895 minDCF