Experiment 1
Binary classification with VGGish model architecture. Logs and metrics have been implemented with tensorboard and pytorch.
Parameters
basic setting parameters
clip_length: 1.0 # [sec]
preprocessing parameters
sample_rate: 22050
hop_length: 512
n_fft: 1024
n_mels: 64
Training parameters
number of audio samples: 22050
learning rate: 0.001
batch size: 10
number of epochs: 10
number of samples: 60
balanced dataset: True
random clip cut: False
Classes
Labels
grime
jazz
Class distribution
Class | Class ID | Samples |
---|---|---|
grime | 0 | 30 |
jazz | 1 | 30 |
Results
Logs
...
Epoch: 10 [30/60 (50%)] Loss: 0.591532 Accuracy: 73.33333333333333%
Epoch: 10 [60/60 (100%)] Loss: 0.585686 Accuracy: 70.0%
[Class: grime] accuracy: 86.7 %
[Class: jazz] accuracy: 56.7 %
Epoch: [10/10] Loss: 0.588609 Accuracy: 71.66666666666667%
[[=============================================================================================]]
Metrics
Cross entropy
![](/_next/static/media/Jan25_22-37-21-loss.931022ae.png)
Accuracy
![](/_next/static/media/acc-Jan25_22-37-21.a1b77028.png)
Loss/Accuracy per batch
![](/_next/static/media/loss-acc-batch.3d3ace10.png)
Accuracy per class
![](/_next/static/media/class-Jan25_22-37-21.3a2aac16.png)
Conv4 Layer
![](/_next/static/media/conv-Jan25_22-37-21.82f37edc.png)
Linear layer
![](/_next/static/media/linear-Jan25_22-37-21.c0c02987.png)
Confusion Matrix
![](/_next/static/media/cf-Jan25_22-37-21.e53d731a.png)