Merge pull request #10 from harritaylor/dev

Dev merge

README.md

@@ -1,35 +1,27 @@
-# Torch VGGish
-A PyTorch port of [VGGish](https://github.com/tensorflow/models/tree/master/research/audioset)<sup>1</sup>, 
+# VGGish
+A `torch`-compatible port of [VGGish](https://github.com/tensorflow/models/tree/master/research/audioset)<sup>[1]</sup>, 
 a feature embedding frontend for audio classification models. The weights are ported directly from the tensorflow model, so embeddings created using `torchvggish` will be identical.
 
-## Quick start
-There are two options: you can install the last stable version from pypi, or clone this repo and install.
-```shell script
-# optional: create virtual env
-cd torchvggish && python3 -m venv .env
-source activate .env/bin/activate
 
-pip install -i https://test.pypi.org/simple/ torchvggish==0.1
-
-# OR get the latest version
-git clone [email protected]:harritaylor/torchvggish.git
-pip install -r requirements.txt
-```
 ## Usage
-Barebones example of creating embeddings from an `example` wav file:
+
 ```python
-from torchvggish import vggish, vggish_input
+import torch
+
+model = torch.hub.load('harritaylor/torchvggish', 'vggish')
+model.eval()
 
-# Initialise model and download weights
-embedding_model = vggish()
-embedding_model.eval()
+# Download an example audio file
+import urllib
+url, filename = ("http://soundbible.com/grab.php?id=1698&type=wav", "bus_chatter.wav")
+try: urllib.URLopener().retrieve(url, filename)
+except: urllib.request.urlretrieve(url, filename)
 
-example = vggish_input.wavfile_to_examples("example.wav")
-embeddings = embedding_model.forward(example)
+model.forward(filename)
 ```
 
 <hr>
-1.  S. Hershey et al., ‘CNN Architectures for Large-Scale Audio Classification’,\
+[1]  S. Hershey et al., ‘CNN Architectures for Large-Scale Audio Classification’,\
     in International Conference on Acoustics, Speech and Signal Processing (ICASSP),2017\
     Available: https://arxiv.org/abs/1609.09430, https://ai.google/research/pubs/pub45611
 

hubconf.py

@@ -0,0 +1,15 @@
+dependencies = ['torch', 'numpy', 'resampy', 'soundfile']
+
+from torchvggish.vggish import VGGish
+
+model_urls = {
+    'vggish': 'https://github.com/harritaylor/torchvggish/'
+              'releases/download/v0.1/vggish-10086976.pth',
+    'pca': 'https://github.com/harritaylor/torchvggish/'
+           'releases/download/v0.1/vggish_pca_params-970ea276.pth'
+}
+
+
+def vggish(**kwargs):
+    model = VGGish(urls=model_urls, **kwargs)
+    return model