Update README.md
Browse files
README.md
CHANGED
|
@@ -40,11 +40,12 @@ model-index:
|
|
| 40 |
name: DIHARD3-eval
|
| 41 |
type: dihard3-eval-1to4spks
|
| 42 |
config: with_overlap_collar_0.0s
|
|
|
|
| 43 |
split: eval
|
| 44 |
metrics:
|
| 45 |
- name: Test DER
|
| 46 |
type: der
|
| 47 |
-
value:
|
| 48 |
- task:
|
| 49 |
name: Speaker Diarization
|
| 50 |
type: speaker-diarization-with-post-processing
|
|
@@ -52,11 +53,12 @@ model-index:
|
|
| 52 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 53 |
type: CALLHOME-part2-2spk
|
| 54 |
config: with_overlap_collar_0.25s
|
|
|
|
| 55 |
split: part2-2spk
|
| 56 |
metrics:
|
| 57 |
- name: Test DER
|
| 58 |
type: der
|
| 59 |
-
value:
|
| 60 |
- task:
|
| 61 |
name: Speaker Diarization
|
| 62 |
type: speaker-diarization-with-post-processing
|
|
@@ -64,11 +66,12 @@ model-index:
|
|
| 64 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 65 |
type: CALLHOME-part2-3spk
|
| 66 |
config: with_overlap_collar_0.25s
|
|
|
|
| 67 |
split: part2-3spk
|
| 68 |
metrics:
|
| 69 |
- name: Test DER
|
| 70 |
type: der
|
| 71 |
-
value:
|
| 72 |
- task:
|
| 73 |
name: Speaker Diarization
|
| 74 |
type: speaker-diarization-with-post-processing
|
|
@@ -76,11 +79,12 @@ model-index:
|
|
| 76 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 77 |
type: CALLHOME-part2-4spk
|
| 78 |
config: with_overlap_collar_0.25s
|
|
|
|
| 79 |
split: part2-4spk
|
| 80 |
metrics:
|
| 81 |
- name: Test DER
|
| 82 |
type: der
|
| 83 |
-
value: 12.
|
| 84 |
- task:
|
| 85 |
name: Speaker Diarization
|
| 86 |
type: speaker-diarization-with-post-processing
|
|
@@ -88,11 +92,12 @@ model-index:
|
|
| 88 |
name: call_home_american_english_speech
|
| 89 |
type: CHAES_2spk_109sessions
|
| 90 |
config: with_overlap_collar_0.25s
|
|
|
|
| 91 |
split: ch109
|
| 92 |
metrics:
|
| 93 |
- name: Test DER
|
| 94 |
type: der
|
| 95 |
-
value:
|
| 96 |
metrics:
|
| 97 |
- der
|
| 98 |
pipeline_tag: audio-classification
|
|
@@ -277,9 +282,9 @@ Data collection methods vary across individual datasets. For example, the above
|
|
| 277 |
|
| 278 |
| **Dataset** | **DIHARD3-Eval <= 4spk** | **CALLHOME-2spk part2** | **CALLHOME-3spk part2** | **CALLHOME-4spk part2** | **CH109** |
|
| 279 |
|:------------------------------------------------------------------------------|:--------------------------:|:------------------------:|:------------------------:|:------------------------:|:------------------:|
|
| 280 |
-
| DER **
|
| 281 |
-
| DER **
|
| 282 |
-
| DER **
|
| 283 |
|
| 284 |
* "IBL" stands for Input Buffer Latency which is identical to chunk length in the streaming implementation.
|
| 285 |
|
|
|
|
| 40 |
name: DIHARD3-eval
|
| 41 |
type: dihard3-eval-1to4spks
|
| 42 |
config: with_overlap_collar_0.0s
|
| 43 |
+
input_buffer_lenght: 1.04s
|
| 44 |
split: eval
|
| 45 |
metrics:
|
| 46 |
- name: Test DER
|
| 47 |
type: der
|
| 48 |
+
value: 13.32
|
| 49 |
- task:
|
| 50 |
name: Speaker Diarization
|
| 51 |
type: speaker-diarization-with-post-processing
|
|
|
|
| 53 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 54 |
type: CALLHOME-part2-2spk
|
| 55 |
config: with_overlap_collar_0.25s
|
| 56 |
+
input_buffer_lenght: 1.04s
|
| 57 |
split: part2-2spk
|
| 58 |
metrics:
|
| 59 |
- name: Test DER
|
| 60 |
type: der
|
| 61 |
+
value: 6.43
|
| 62 |
- task:
|
| 63 |
name: Speaker Diarization
|
| 64 |
type: speaker-diarization-with-post-processing
|
|
|
|
| 66 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 67 |
type: CALLHOME-part2-3spk
|
| 68 |
config: with_overlap_collar_0.25s
|
| 69 |
+
input_buffer_lenght: 1.04s
|
| 70 |
split: part2-3spk
|
| 71 |
metrics:
|
| 72 |
- name: Test DER
|
| 73 |
type: der
|
| 74 |
+
value: 10.26
|
| 75 |
- task:
|
| 76 |
name: Speaker Diarization
|
| 77 |
type: speaker-diarization-with-post-processing
|
|
|
|
| 79 |
name: CALLHOME (NIST-SRE-2000 Disc8)
|
| 80 |
type: CALLHOME-part2-4spk
|
| 81 |
config: with_overlap_collar_0.25s
|
| 82 |
+
input_buffer_lenght: 1.04s
|
| 83 |
split: part2-4spk
|
| 84 |
metrics:
|
| 85 |
- name: Test DER
|
| 86 |
type: der
|
| 87 |
+
value: 12.40
|
| 88 |
- task:
|
| 89 |
name: Speaker Diarization
|
| 90 |
type: speaker-diarization-with-post-processing
|
|
|
|
| 92 |
name: call_home_american_english_speech
|
| 93 |
type: CHAES_2spk_109sessions
|
| 94 |
config: with_overlap_collar_0.25s
|
| 95 |
+
input_buffer_lenght: 1.04s
|
| 96 |
split: ch109
|
| 97 |
metrics:
|
| 98 |
- name: Test DER
|
| 99 |
type: der
|
| 100 |
+
value: 5.09
|
| 101 |
metrics:
|
| 102 |
- der
|
| 103 |
pipeline_tag: audio-classification
|
|
|
|
| 282 |
|
| 283 |
| **Dataset** | **DIHARD3-Eval <= 4spk** | **CALLHOME-2spk part2** | **CALLHOME-3spk part2** | **CALLHOME-4spk part2** | **CH109** |
|
| 284 |
|:------------------------------------------------------------------------------|:--------------------------:|:------------------------:|:------------------------:|:------------------------:|:------------------:|
|
| 285 |
+
| DER **Input Buffer Length: 1.04s** | 14.57 | 7.35 | 11.57 | 13.83 | 5.59 |
|
| 286 |
+
| DER **Input Buffer Length: 1.04s + DH3-dev Opt. PP** | **_13.32_** | - | - | - | - |
|
| 287 |
+
| DER **Input Buffer Length: 1.04s + CallHome-part1 Opt. PP** | - | **_6.43_** | **_10.26_** | **_12.40_** | **_5.09_** |
|
| 288 |
|
| 289 |
* "IBL" stands for Input Buffer Latency which is identical to chunk length in the streaming implementation.
|
| 290 |
|