During first lesson of Practical Deep Learning for Coders course, Jeremy had mentioned how using simple computer vision model by being a bit creative we can build a state of the art model to classify audio with same image classification model. I was curious on how I can train an music classifier, as I have never worked on audio data before.


[You can find how I trained this music genre classification using fast.ai](https://kurianbenoy.com/ml-blog/fastai/fastbook/2022/05/01/AudioCNNDemo.html).

## Dataset

1. [The competition data](https://www.kaggle.com/competitions/kaggle-pog-series-s01e02/data)
2. [Image data generated from converting audio to melspectograms in form of images](https://www.kaggle.com/datasets/dienhoa/music-genre-spectrogram-pogchamps)


## Training

Fast.ai was used to train this classifier with a ResNet50 vision learner for 10 epochs.

| epoch	| train_loss	| valid_loss	| error_rate	| time  |
|-------|---------------|---------------|---------------|-------|
|0  |	2.312176 |	1.843815 |	0.558654 |	02:07 |
|1  |	2.102361 |	1.719162 |	0.539061 |	02:08 |
|2  |	1.867139 |	1.623988 |	0.527003 |	02:08 |
|3  |	1.710557 |	1.527913 |	0.507661 |	02:07 |
|4  |	1.629478 |	1.456836 |	0.479779 |	02:05 |
|5  |	1.519305 |	1.433036 |	0.474253 |	02:05 |
|6  |	1.457465 |	1.379757 |	0.464456 |	02:05 |
|7  |	1.396283 |	1.369344 |	0.457925 |	02:05 |
|8  |	1.359388 |	1.367973 |	0.453655 |	02:05 |
|9  |	1.364363 |	1.368887 |	0.456167 |	02:04 |



| epoch	| train_loss	| valid_loss	| error_rate	| time  |
|-------|---------------|---------------|---------------|-------|
| 0	| 1.358123	| 1.100139	| 0.288713	| 05:14 |
| 1	| 1.129988	| 0.985213	| 0.260693	| 05:12 |
| 2	| 0.964907	| 0.909715	| 0.241337	| 05:17 |
| 3	| 0.804738	| 0.843515	| 0.222475	| 05:19 |
| 4	| 0.638846	| 0.795957	| 0.205347	| 05:16 |
| 5	| 0.475434	| 0.750069	| 0.192673	| 05:15 |
| 6	| 0.345060	| 0.742432	| 0.185198	| 05:12 |
| 7	| 0.247938	| 0.728758	| 0.177624	| 05:12 |
| 8	| 0.214708	| 0.727486	| 0.177871	| 05:11 |

## Examples

The example images provided in the demo are from the validation data from Kaggle competition data, which was not used during training.