jefflai108 / Contrastive Predictive Coding Pytorch
Licence: mit
Contrastive Predictive Coding for Automatic Speaker Verification
Stars: ✭ 223
Programming Languages
python
139335 projects - #7 most used programming language
Projects that are alternatives of or similar to Contrastive Predictive Coding Pytorch
Contrastive Predictive Coding
Keras implementation of Representation Learning with Contrastive Predictive Coding
Stars: ✭ 369 (+65.47%)
Mutual labels: unsupervised-learning, representation-learning, predictive-modeling
Lemniscate.pytorch
Unsupervised Feature Learning via Non-parametric Instance Discrimination
Stars: ✭ 532 (+138.57%)
Mutual labels: unsupervised-learning, representation-learning
Awesome Vaes
A curated list of awesome work on VAEs, disentanglement, representation learning, and generative models.
Stars: ✭ 418 (+87.44%)
Mutual labels: unsupervised-learning, representation-learning
Pytorch Byol
PyTorch implementation of Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning
Stars: ✭ 213 (-4.48%)
Mutual labels: unsupervised-learning, representation-learning
Pytorch Cortexnet
PyTorch implementation of the CortexNet predictive model
Stars: ✭ 349 (+56.5%)
Mutual labels: unsupervised-learning, predictive-modeling
Disentangling Vae
Experiments for understanding disentanglement in VAE latent representations
Stars: ✭ 398 (+78.48%)
Mutual labels: unsupervised-learning, representation-learning
Simclr
PyTorch implementation of SimCLR: A Simple Framework for Contrastive Learning of Visual Representations
Stars: ✭ 750 (+236.32%)
Mutual labels: unsupervised-learning, representation-learning
ladder-vae-pytorch
Ladder Variational Autoencoders (LVAE) in PyTorch
Stars: ✭ 59 (-73.54%)
Mutual labels: representation-learning, unsupervised-learning
Transferlearning
Transfer learning / domain adaptation / domain generalization / multi-task learning etc. Papers, codes, datasets, applications, tutorials.-迁移学习
Stars: ✭ 8,481 (+3703.14%)
Mutual labels: representation-learning, unsupervised-learning
Self Supervised Learning Overview
📜 Self-Supervised Learning from Images: Up-to-date reading list.
Stars: ✭ 73 (-67.26%)
Mutual labels: unsupervised-learning, representation-learning
Pointglr
Global-Local Bidirectional Reasoning for Unsupervised Representation Learning of 3D Point Clouds (CVPR 2020)
Stars: ✭ 86 (-61.43%)
Mutual labels: unsupervised-learning, representation-learning
Simclr
PyTorch implementation of SimCLR: A Simple Framework for Contrastive Learning of Visual Representations by T. Chen et al.
Stars: ✭ 293 (+31.39%)
Mutual labels: unsupervised-learning, representation-learning
ML2017FALL
Machine Learning (EE 5184) in NTU
Stars: ✭ 66 (-70.4%)
Mutual labels: predictive-modeling, unsupervised-learning
srVAE
VAE with RealNVP prior and Super-Resolution VAE in PyTorch. Code release for https://arxiv.org/abs/2006.05218.
Stars: ✭ 56 (-74.89%)
Mutual labels: representation-learning, unsupervised-learning
Variational Ladder Autoencoder
Implementation of VLAE
Stars: ✭ 196 (-12.11%)
Mutual labels: unsupervised-learning, representation-learning
Unsupervised Classification
SCAN: Learning to Classify Images without Labels (ECCV 2020), incl. SimCLR.
Stars: ✭ 605 (+171.3%)
Mutual labels: unsupervised-learning, representation-learning
SimCLR
Pytorch implementation of "A Simple Framework for Contrastive Learning of Visual Representations"
Stars: ✭ 65 (-70.85%)
Mutual labels: representation-learning, unsupervised-learning
awesome-contrastive-self-supervised-learning
A comprehensive list of awesome contrastive self-supervised learning papers.
Stars: ✭ 748 (+235.43%)
Mutual labels: representation-learning, unsupervised-learning
Bagofconcepts
Python implementation of bag-of-concepts
Stars: ✭ 18 (-91.93%)
Mutual labels: unsupervised-learning, representation-learning
Autoregressive Predictive Coding
Autoregressive Predictive Coding: An unsupervised autoregressive model for speech representation learning
Stars: ✭ 138 (-38.12%)
Mutual labels: unsupervised-learning, representation-learning
Contrastive-Predictive-Coding-PyTorch
This repository contains (PyTorch) code to reproduce the core results for:
- Representation Learning with Contrastive Predictive Coding
- Contrastive Predictive Coding Based Feature for Automatic Speaker Verification
If you find the code useful, please cite
@article{lai2019contrastive,
title={Contrastive Predictive Coding Based Feature for Automatic Speaker Verification},
author={Lai, Cheng-I},
journal={arXiv preprint arXiv:1904.01575},
year={2019}
}
Getting Started
./src/model/model.py contains the CPC models implementation, ./src/main.py is the code for training the CPC models, ./src/spk_class.py trains a NN speaker classifier, ./ivector/ contains the scripts for running an i-vectors speaker verification system.
An example of CPC and speaker classifier training can be found at
./run.sh
CPC Models
CDCK2: base model from the paper 'Representation Learning with Contrastive Predictive Coding'.
CDCK5: CDCK2 with a different decoder.
CDCK6: CDCK2 with a shared encoder and double decoders.
Experimental Results
A. CPC Model Training
| CPC model ID | number of epoch | model size | dev NCE loss | dev acc. |
|---|---|---|---|---|
| CDCK2 | 60 | 7.42M | 1.6427 | 26.42 |
| CDCK5 | 60 | 5.58M | 1.7818 | 22.48 |
| CDCK6 | 30 | 7.33M | 1.6484 | 28.24 |
B. Speaker Verificaiton on LibriSpeech test-clean-100 (Average Pooling)
Note: 1st trial list and 2nd trial list
| Feature | Feature Dim | Summarization | LDA Dim | 1st EER | 2nd EER |
|---|---|---|---|---|---|
| MFCC | 24 | average pooling | 24 | 9.211 | 13.48 |
| CDCK2 | 256 | average pooling | 200 | 5.887 | 11.1 |
| CDCK5 | 40 | average pooling | 40 | 7.508 | 12.25 |
| CDCK6 | 256 | average pooling | 200 | 6.809 | 12.73 |
C. CPC applied with PCA
| Feature w PCA | Original Feature | PCA Dim | PCA Variance Ratio |
|---|---|---|---|
| CDCK2-36 | CDCK2 | 36 | 76.76 |
| CDCK2-60 | CDCK2 | 60 | 87.40 |
| CDCK5-24 | CDCK5 | 24 | 93.39 |
D. Speaker Verificaiton on LibriSpeech test-clean-100 (i-vectors)
| Feature | Feature Dim | Summarization | 1st EER | 2nd EER |
|---|---|---|---|---|
| MFCC | 24 | i-vectors | 5.518 | 8.157 |
| CDCK2-60 | 60 | i-vectors | 5.351 | 9.753 |
| CDCK5-24 | 24 | i-vectors | 4.911 | 8.901 |
| CDCK6-60 | 60 | i-vectors | 5.228 | 9.009 |
| MFCC + CDCK2-36 | 60 | i-vectors | 3.62 | 6.898 |
| MFCC + CDCK5-24 | 48 | i-vectors | 3.712 | 6.962 |
| MFCC + CDCK6-36 | 60 | i-vectors | 3.691 | 6.765 |
E. DET Curves of CPC and MFCC Fusion for i-vectors Speaker Verification
Authors
Cheng-I Lai.
If you encouter any problem, feel free to contact me.
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].