Open IvanEvan opened 6 years ago
The code for the CNN has not been released because I was still testing which approach I should use. This project takes up very little of my time, and I have not been working on it recently. I've done some testing with other projects on whether to use transfer learning or to train a CNN from scratch. I've decided to go with transfer learning and to retrain the last convolutional block of a VGG19 network pre-trained on ImageNet. I'll work on it a little more, but the accuracy for this model will be fairly low as so much voice data is needed to properly identify who's talking.
Using a classical DCNN might not even work, as audio data differs drastically between sessions. Who knows...
@prmelehan Thanks for your reply! I found another project , its CNN architecture maybe useful. Here is link : https://github.com/astorfi/3D-convolutional-speaker-recognition
thank you for this training phase.it looks amazing ! can i have code for testing ? thanks in advance
Hi, I am very interested in your project, so I downloaded your code and want implement it on my computer. But I found your code just have recording.py. There is not have the code about CNN and other necessary code. So, can you complete your project and update other code? Many thanks in adavance.