Robustness small changes

Overall scores

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

Overall Score

66.0% passed tests (33 passed / 17 failed).

78.0% passed tests (39 passed / 11 failed).

62.0% passed tests (31 passed / 19 failed).

56.0% passed tests (28 passed / 22 failed).

Percentage Unchanged Predictions Additive Tone

Threshold: 0.95

Data

Percent Unchanged Pred Additive Tone

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.96

0.98

0.97

0.95

emovo-1.2.1-emotion.test

0.93

0.96

0.95

0.96

iemocap-2.3.0-emotion.categories.test.gold_standard

0.92

0.97

0.96

0.95

meld-1.3.1-emotion.categories.test.gold_standard

0.97

0.99

0.98

0.97

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.95

0.98

0.94

0.96

mean

0.95

0.98

0.96

0.96

Percentage Unchanged Predictions Append Zeros

Threshold: 0.95

Data

Percent Unchanged Pred Append Zeros

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.99

0.99

0.94

0.95

emovo-1.2.1-emotion.test

0.98

0.99

0.98

0.98

iemocap-2.3.0-emotion.categories.test.gold_standard

0.98

0.99

0.96

0.96

meld-1.3.1-emotion.categories.test.gold_standard

0.98

0.98

0.95

0.95

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.99

0.99

0.96

0.98

mean

0.98

0.99

0.96

0.96

Percentage Unchanged Predictions Clip

Threshold: 0.95

Data

Percent Unchanged Pred Clip

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.98

0.98

0.98

0.97

emovo-1.2.1-emotion.test

0.99

0.99

0.99

0.99

iemocap-2.3.0-emotion.categories.test.gold_standard

0.98

0.98

0.97

0.98

meld-1.3.1-emotion.categories.test.gold_standard

0.98

0.98

0.98

0.98

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.99

0.99

0.99

0.99

mean

0.98

0.98

0.98

0.98

Percentage Unchanged Predictions Crop Beginning

Threshold: 0.95

Data

Percent Unchanged Pred Crop Beginning

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.92

0.95

0.94

0.92

emovo-1.2.1-emotion.test

0.91

0.95

0.95

0.93

iemocap-2.3.0-emotion.categories.test.gold_standard

0.93

0.95

0.95

0.92

meld-1.3.1-emotion.categories.test.gold_standard

0.90

0.92

0.93

0.89

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.96

0.97

0.95

0.95

mean

0.92

0.95

0.94

0.92

Percentage Unchanged Predictions Crop End

Threshold: 0.95

Data

Percent Unchanged Pred Crop End

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.99

0.99

0.99

0.98

emovo-1.2.1-emotion.test

0.98

0.99

1.00

0.99

iemocap-2.3.0-emotion.categories.test.gold_standard

0.98

0.99

0.99

0.98

meld-1.3.1-emotion.categories.test.gold_standard

0.97

0.96

0.96

0.96

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.98

0.99

0.99

0.98

mean

0.98

0.98

0.99

0.98

Percentage Unchanged Predictions Gain

Threshold: 0.95

Data

Percent Unchanged Pred Gain

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

1.00

1.00

0.99

1.00

emovo-1.2.1-emotion.test

1.00

1.00

0.98

1.00

iemocap-2.3.0-emotion.categories.test.gold_standard

0.99

1.00

0.99

0.99

meld-1.3.1-emotion.categories.test.gold_standard

1.00

1.00

0.98

1.00

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

1.00

1.00

0.98

1.00

mean

1.00

1.00

0.98

1.00

Percentage Unchanged Predictions Highpass Filter

Threshold: 0.95

Data

Percent Unchanged Pred Highpass Filter

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.97

0.97

0.97

0.96

emovo-1.2.1-emotion.test

0.94

0.98

0.99

0.96

iemocap-2.3.0-emotion.categories.test.gold_standard

0.96

0.98

0.98

0.95

meld-1.3.1-emotion.categories.test.gold_standard

0.96

0.97

0.97

0.95

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.98

0.99

0.97

0.98

mean

0.96

0.98

0.98

0.96

Percentage Unchanged Predictions Lowpass Filter

Threshold: 0.95

Data

Percent Unchanged Pred Lowpass Filter

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.97

0.98

0.94

0.94

emovo-1.2.1-emotion.test

0.97

0.99

0.99

0.99

iemocap-2.3.0-emotion.categories.test.gold_standard

0.99

0.99

0.99

0.98

meld-1.3.1-emotion.categories.test.gold_standard

0.97

0.98

0.98

0.95

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.98

0.99

0.96

0.98

mean

0.98

0.99

0.97

0.97

Percentage Unchanged Predictions Prepend Zeros

Threshold: 0.95

Data

Percent Unchanged Pred Prepend Zeros

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.93

0.94

0.90

0.91

emovo-1.2.1-emotion.test

0.89

0.95

0.95

0.92

iemocap-2.3.0-emotion.categories.test.gold_standard

0.93

0.95

0.93

0.92

meld-1.3.1-emotion.categories.test.gold_standard

0.90

0.92

0.92

0.87

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.95

0.97

0.94

0.95

mean

0.92

0.95

0.93

0.91

Percentage Unchanged Predictions White Noise

Threshold: 0.95

Data

Percent Unchanged Pred White Noise

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

crema-d-1.2.0-emotion.categories.test.gold_standard

0.94

0.96

0.96

0.94

emovo-1.2.1-emotion.test

0.87

0.93

0.91

0.91

iemocap-2.3.0-emotion.categories.test.gold_standard

0.87

0.95

0.92

0.93

meld-1.3.1-emotion.categories.test.gold_standard

0.97

0.97

0.95

0.96

msppodcast-2.6.0-emotion.categories.test-1.gold_standard

0.92

0.95

0.92

0.95

mean

0.91

0.95

0.93

0.94

Visualization

w2v2-L-cat

hubert-L-cat

wavlm-cat

data2vec-cat

../../../_images/visualization_crema-d-1.2.0-emotion.categories.test.gold_standard38.png
../../../_images/visualization_crema-d-1.2.0-emotion.categories.test.gold_standard63.png
../../../_images/visualization_crema-d-1.2.0-emotion.categories.test.gold_standard64.png
../../../_images/visualization_crema-d-1.2.0-emotion.categories.test.gold_standard65.png
../../../_images/visualization_emovo-1.2.1-emotion.test38.png
../../../_images/visualization_emovo-1.2.1-emotion.test63.png
../../../_images/visualization_emovo-1.2.1-emotion.test64.png
../../../_images/visualization_emovo-1.2.1-emotion.test65.png
../../../_images/visualization_iemocap-2.3.0-emotion.categories.test.gold_standard38.png
../../../_images/visualization_iemocap-2.3.0-emotion.categories.test.gold_standard63.png
../../../_images/visualization_iemocap-2.3.0-emotion.categories.test.gold_standard64.png
../../../_images/visualization_iemocap-2.3.0-emotion.categories.test.gold_standard65.png
../../../_images/visualization_meld-1.3.1-emotion.categories.test.gold_standard48.png
../../../_images/visualization_meld-1.3.1-emotion.categories.test.gold_standard85.png
../../../_images/visualization_meld-1.3.1-emotion.categories.test.gold_standard86.png
../../../_images/visualization_meld-1.3.1-emotion.categories.test.gold_standard87.png
../../../_images/visualization_msppodcast-2.6.0-emotion.categories.test-1.gold_standard26.png
../../../_images/visualization_msppodcast-2.6.0-emotion.categories.test-1.gold_standard63.png
../../../_images/visualization_msppodcast-2.6.0-emotion.categories.test-1.gold_standard64.png
../../../_images/visualization_msppodcast-2.6.0-emotion.categories.test-1.gold_standard65.png