Artem Gorlanov
commited on
Commit
•
d7280b3
1
Parent(s):
54d944a
fix
Browse files
README.md
CHANGED
|
@@ -37,9 +37,7 @@ docker run -it -p 7860:7860 --platform=linux/amd64 \
|
|
| 37 |
| cpu 2v CPU huggingece | 453.8 s/it |
|
| 38 |
| gpu tesla v100 | 8.23 s/it |
|
| 39 |
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
I know a lot of methods for this task:
|
| 44 |
+ separation: using separation models(need longtime train and finetune)
|
| 45 |
+ diarization
|
|
@@ -52,10 +50,10 @@ I know a lot of methods for this task:
|
|
| 52 |
For this task i used speaker_embedding+Clustering unknowing numbers of speakers
|
| 53 |
|
| 54 |
|
| 55 |
-
How i can improve
|
| 56 |
-
+ preprocessing
|
| 57 |
+ estimate SNR(signal noise rate) and if input clean dont use denoising
|
| 58 |
-
+ train:
|
| 59 |
+ custom speaker recognition model
|
| 60 |
+ custom overlap speech detector
|
| 61 |
+ custom speech separation model:
|
|
|
|
| 37 |
| cpu 2v CPU huggingece | 453.8 s/it |
|
| 38 |
| gpu tesla v100 | 8.23 s/it |
|
| 39 |
|
| 40 |
+
# Approaches
|
|
|
|
|
|
|
| 41 |
I know a lot of methods for this task:
|
| 42 |
+ separation: using separation models(need longtime train and finetune)
|
| 43 |
+ diarization
|
|
|
|
| 50 |
For this task i used speaker_embedding+Clustering unknowing numbers of speakers
|
| 51 |
|
| 52 |
|
| 53 |
+
# How i can improve:
|
| 54 |
+
+ Fix preprocessing
|
| 55 |
+ estimate SNR(signal noise rate) and if input clean dont use denoising
|
| 56 |
+
+ Add train:
|
| 57 |
+ custom speaker recognition model
|
| 58 |
+ custom overlap speech detector
|
| 59 |
+ custom speech separation model:
|