Trial Set

CN-Celeb Genre

Uses CN-Celeb metadata to build a genre-shift supplement outside the official trial file.

Notice: This is not the original official SV trial list.

Source dataset
CN-Celeb
Trials
440000
Protocol source
Self-constructed supplement
Source split
CN-Celeb prepared evaluation pool
cnceleb_genre Genre shift
Source Dataset

About CN-Celeb

CN-Celeb was built to study Chinese speaker recognition in unconstrained media conditions. The original collection contains more than 130,000 utterances from 1,000 celebrities across 11 real-world genres; the later CN-Celeb2 release expands the multi-genre collection to 3,000 speakers. The main OpenSVBench entry uses the upstream evaluation protocol, while the genre and short-duration entries are separate supplements.

Language
Chinese
Original CN-Celeb
1,000 speakers; 130,000+ utterances
Genres
11 real-world media genres
Recording
In-the-wild media speech
Direct Ranking

Systems on this trial set

26 reviewed systems currently have valid results.

Rank System Submitted by Trial-Set EER ↓ Trial-Set minDCF ↓
#1 w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2
w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2
JinghanPeng 13.2400 0.461908
#2 w2vbert2.0+ MHFA+uncertainty
w2vbert2.0+ MHFA+uncertainty
abc 13.5700 0.820707
#3 iic/eres2netv2-zh
eresv2_zh
Public checkpoint 14.2725 0.527525
#4 wespeaker/res34-cnceleb
wespeaker_r34
Public checkpoint 15.6955 0.591628
#5 iic/campplus
campplus
Public checkpoint 16.8815 0.610655
#6 W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2)
Wav2Vec2-BERT
pengjy 19.0005 0.737320
#7 W2V-BERT 2.0 + Adapter-MFA (cross-device consistency FT + weight soup)
ActionPower VoicePrint
ActionPower 19.4650 0.626777
#8 MECT-b2-czur-finetuned
MECT-b2-czur-finetuned
czur 19.6400 0.649620
#9 Agora Voiceprint (ResNet293)
agora_vp
agora 21.9825 0.674857
#10 Dual-LoRA W2V-BERT2-Adapter-MFA (TidyVoice)
duallora-w2vbert2-tidyvoice-full-core
sgqt 22.2000 0.654345
#11 AntSpeaker/MECT-B2
antspeaker-mect-b2-vb2-full-core
Public checkpoint 23.4125 0.648552
#12 palabraai/redimnet2-b6-vb2+vox2
redimnet2-b6-vb2vox2-lm-full-core
Public checkpoint 23.5245 0.640800
#13 wespeaker/samresnet100
vb2-sam100
Public checkpoint 25.1200 0.674765
#14 palabraai/redimnet2-b6
redimnet2
Public checkpoint 25.4550 0.740052
#15 W2V-BERT 2.0 + LoRA_Adapter_MFA (VoxCeleb2 dev + VoxBlink2)
vb2-w2vbert
Public checkpoint 25.9200 0.656577
#16 wespeaker/res152-voxceleb
wespeaker_resnet152
Public checkpoint 26.2472 0.806172
#17 wespeaker/res34-voxceleb
wespeaker_resnet34_voxceleb
Public checkpoint 26.5125 0.822805
#18 wespeaker/res293-voxceleb
wespeaker_resnet293
Public checkpoint 26.8575 0.811767
#19 iic/eres2net-en
eres_en
Public checkpoint 27.7395 0.802170
#20 iic/eres2net-large-3dspeaker
eres2net_large_3dspeaker
Public checkpoint 28.4850 0.888960
#21 speechbrain/ecapa
ecapa
Public checkpoint 29.5690 0.883898
#22 wespeaker/campplus-voxceleb
campplus_wespeaker
Public checkpoint 29.6497 0.829445
#23 wespeaker/ecapa512-voxceleb
wespeaker_ecapa512
Public checkpoint 30.8550 0.849922
#24 wespeaker/ecapa1024-voxceleb
wespeaker_ecapa1024
Public checkpoint 31.1025 0.910323
#25 microsoft/unispeech-sat-base-plus-sv
unispeech_sat_base_plus
Public checkpoint 37.0343 0.985940
#26 microsoft/wavlm-base-plus-sv
wavlm_base
Public checkpoint 38.3800 0.975515