Small datasets to test novel audio algorithms, heavily influenced by imagenette
The ESC-50 dataset is a labeled collection of 2000 environmental audio recordings suitable for benchmarking methods of environmental sound classification.
The dataset consists of 5-second-long recordings organized into 50 semantical classes (with 40 examples per class) loosely arranged into 5 major categories:
ESC-50 is a really nice starting dataset as it is especially clean (fixed-length, hand-labeled, single sample-rate) and well maintained. Many thanks to Karol Piczak for maintaining a really great Github Repo based around this dataset.
If you are already using the fastaudio library, you can download and access these quickly with commands like:
path = untar_data(URLs.ESC50)where path now stores the destination to ESC-50.
Generally you'll see +/- 1% differences from run to run since it's quite a small validation set. So please only send in contributions that are higher than the reported accuracy >80% of the time. Here's the rules:
- No inference time tricks, e.g. no: TTA
- Must be one of the split/#epoch combinations listed in the table
- If you have the resources to do so, try to get an average of 5 runs, to get a stable comparison. Use the "# Runs" column to include this
- In the URL column include a link to a notebook, blog post, gist, or similar which explains what you did to get your result, and includes the code you used (or a link to it), including the exact commit, so that others can reproduce your result.
- For this leaderbord, use only the data corresponding to Fold 1 to validate, and train on the other 4 folds.
| Epochs | URL | Accuracy | # Runs |
|---|---|---|---|
| 80 | fastaudio baseline w/ mixup | 78.25% | 1 |
| 20 | fastaudio baseline | 66.64% | 5, mean |
| 10 | fastaudio baseline | 62.69% | 5, mean |
-
For this leaderbord, do 5-fold cross validation using the splits defined in the metadata, and report the mean accuracy.
-
The results here are also comparable with the official dataset leaderbods
| Epochs | URL | Accuracy | # Runs |
|---|---|---|---|
| 80 | no entries yet | -- | -- |
| 20 | fastaudio baseline | 67.35% | 1 |
| 10 | fastaudio baseline | 64.94% | 1 |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent