Spaces:
Runtime error
Runtime error
File size: 2,276 Bytes
12c06de 19692ef 31a76f2 12c06de d6a5961 c8f7b53 f50c50e c8f7b53 77e643c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
---
title: VoID Demo from AISF Hackathon
emoji: 😤
colorFrom: blue
colorTo: yellow
sdk: gradio
python_version: 3.9
app_file: gradio_app.py
pinned: false
---
# VoID: Voice Identifier
Classifier to recognize the identity of a speaker
Project was done as a part of the AISF hackathon (www.aisf.co), VoID was a finalist :D
## Details
### Dataset
Custom dataset with a set of 5-7 pure, unaugmented voices from labeled speakers saying the same phrase (similar to a passcode). In our initial version, we had three voices saying the same phrase ("The quick brown fox jumps over the lazy dog"). In our demo, we used three voices saying their own individual name.
For test and training data, we then created hundreds of augmented copies of each pure voice using augmentation layers (as seen in the augmentation notebook under the `/notebooks/` folder). The types of distortions were added in randomly and were even compounded randomly, augmentations such as clipping, noise, reverb, and etc.
If you would like access to the dataset used in this project, feel free to contact me.
### Model Structure
```
CNNetwork(
(conv1): Sequential(
(0): Conv2d(1, 16, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2))
(1): ReLU()
(2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(conv2): Sequential(
(0): Conv2d(16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2))
(1): ReLU()
(2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(conv3): Sequential(
(0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2))
(1): ReLU()
(2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(conv4): Sequential(
(0): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2))
(1): ReLU()
(2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(flatten): Flatten(start_dim=1, end_dim=-1)
(linear): Linear(in_features=21888, out_features=3, bias=True)
(softmax): Softmax(dim=1)
)
```
## Demo
Watch the demo of our model [here](https://www.loom.com/share/a8cb126af7b64ddaaa67c6f00e23f4e9)!
## Running locally
Make sure to create and install the conda dependencies declared in `void.yml`.
|