Hugging Face's logo

mechanicalsea
/
efficient-tdnn

PyTorch
English
embeddings
Speaker
Verification
Identification
NAS
TDNN
Model card Files Community
mechanicalsea commited on
Commit
edfae8a
•
1 Parent(s): 16c4d0d

remove performance without AS-Norm for concise

Browse files
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -27,11 +27,11 @@ The system can be used to extract speaker embeddings with different model size.
27
  It is trained on Voxceleb2 training data using data augmentation.
28
  The model performance on Voxceleb1-test set(Cleaned)/Vox1-O are reported as follows.
29
 
30
- | Supernet Stage | Subnet | MACs (3-second) | Params | EER(%) w/ AS-Norm | EER(%) w/o AS-Norm | minDCF w/ AS-Norm | minDCF w/o AS-Norm |
31
- |:-------------:|:--------------:|:--------------:|:--------------:|:--------------:|:--------------:|:--------------:|:--------------:|
32
- | depth | Base | 1.45G | 5.79M | 0.94 | 1.14 | 0.089 | 0.106 |
33
- | width 1 | Mobile | 570.98M | 2.42M | 1.41 | 1.61 | 0.124 | 0.152 |
34
- | width 2 | Small | 204.07M | 899.20K | 2.20 | 2.33 | 0.219 | 0.241 |
35
 
36
  The details of three subnets are:
37
 
 
27
  It is trained on Voxceleb2 training data using data augmentation.
28
  The model performance on Voxceleb1-test set(Cleaned)/Vox1-O are reported as follows.
29
 
30
+ | Supernet Stage | Subnet | MACs (3s) | Params | EER(%) | minDCF |
31
+ |:-------------:|:--------------:|:--------------:|:--------------:|:--------------:|:--------------:|
32
+ | depth | Base | 1.45G | 5.79M | 0.94 | 0.089 |
33
+ | width 1 | Mobile | 570.98M | 2.42M | 1.41 | 0.124 |
34
+ | width 2 | Small | 204.07M | 899.20K | 2.20 | 0.219 |
35
 
36
  The details of three subnets are:
37