CNN14

79.2% passed tests (350 passed / 92 failed).

Tests overview

Topic

Passed Tests

Correctness consistency

40.4%

Correctness distribution

33.3%

Correctness regression

0.0%

Correctness speaker average

100.0%

Correctness speaker ranking

0.0%

Fairness accent

98.1%

Fairness language

100.0%

Fairness linguistic sentiment

97.9%

Fairness pitch

80.0%

Fairness sex

83.3%

Robustness background noise

41.7%

Robustness low quality phone

0.0%

Robustness recording condition

0.0%

Robustness simulated recording condition

0.0%

Robustness small changes

70.0%

Robustness spectral tilt

12.5%

Model parameters

Entry

Value

data

msppodcast-2.3.0

tasks

arousal, valence, dominance

features

torchaudio.transforms.MelSpectrogram

sampling_rate

16000

quantization

False

output_range

(0, 1)

Model information

Entry

Value

Author

jwagner (onnx export: aderington)

Date

2023-04-06

Name

onnx

Subgroup

ser.adv.audpann.cnn14

Version

1.0.3

Sample

{‘emodb’: {‘version’: ‘1.1.1’, ‘flavor’: {‘bit_depth’: None, ‘channels’: None, ‘format’: ‘wav’, ‘mixdown’: True, ‘sampling_rate’: 16000}, ‘data’: {‘wav/03a01Fa.wav’: [[0.6487703323364258, 0.19668646156787872, 0.681863009929657]], ‘wav/08a01Ab.wav’: [[0.761000394821167, 0.09757468849420547, 0.7346381545066833]]}}}