Sleepychord / Improvedgan Pytorch
Semi-supervised GAN in "Improved Techniques for Training GANs"
Stars: ✭ 228
Programming Languages
python
139335 projects - #7 most used programming language
Projects that are alternatives of or similar to Improvedgan Pytorch
ganbert-pytorch
Enhancing the BERT training with Semi-supervised Generative Adversarial Networks in Pytorch/HuggingFace
Stars: ✭ 60 (-73.68%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
CsiGAN
An implementation for our paper: CsiGAN: Robust Channel State Information-based Activity Recognition with GANs (IEEE Internet of Things Journal, 2019), which is the semi-supervised Generative Adversarial Network (GAN) for Channel State Information (CSI) -based activity recognition.
Stars: ✭ 23 (-89.91%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
ganbert
Enhancing the BERT training with Semi-supervised Generative Adversarial Networks
Stars: ✭ 205 (-10.09%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Ssgan Tensorflow
A Tensorflow implementation of Semi-supervised Learning Generative Adversarial Networks (NIPS 2016: Improved Techniques for Training GANs).
Stars: ✭ 496 (+117.54%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Ali Pytorch
PyTorch implementation of Adversarially Learned Inference (BiGAN).
Stars: ✭ 61 (-73.25%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Accel Brain Code
The purpose of this repository is to make prototypes as case study in the context of proof of concept(PoC) and research and development(R&D) that I have written in my website. The main research topics are Auto-Encoders in relation to the representation learning, the statistical machine learning for energy-based models, adversarial generation networks(GANs), Deep Reinforcement Learning such as Deep Q-Networks, semi-supervised learning, and neural network language model for natural language processing.
Stars: ✭ 166 (-27.19%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
catgan pytorch
Unsupervised and Semi-supervised Learning with Categorical Generative Adversarial Networks
Stars: ✭ 50 (-78.07%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Gans In Action
Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks
Stars: ✭ 748 (+228.07%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Acgan Pytorch
Pytorch implementation of Conditional Image Synthesis with Auxiliary Classifier GANs
Stars: ✭ 57 (-75%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Sparsely Grouped Gan
Code for paper "Sparsely Grouped Multi-task Generative Adversarial Networks for Facial Attribute Manipulation"
Stars: ✭ 68 (-70.18%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Triple Gan
See Triple-GAN-V2 in PyTorch: https://github.com/taufikxu/Triple-GAN
Stars: ✭ 203 (-10.96%)
Mutual labels: generative-adversarial-network, semi-supervised-learning
Ranksrgan
ICCV 2019 (oral) RankSRGAN: Generative Adversarial Networks with Ranker for Image Super-Resolution. PyTorch implementation
Stars: ✭ 213 (-6.58%)
Mutual labels: generative-adversarial-network
Iseebetter
iSeeBetter: Spatio-Temporal Video Super Resolution using Recurrent-Generative Back-Projection Networks | Python3 | PyTorch | GANs | CNNs | ResNets | RNNs | Published in Springer Journal of Computational Visual Media, September 2020, Tsinghua University Press
Stars: ✭ 202 (-11.4%)
Mutual labels: generative-adversarial-network
Conditional Gan
Tensorflow implementation for Conditional Convolutional Adversarial Networks.
Stars: ✭ 202 (-11.4%)
Mutual labels: generative-adversarial-network
Csa Inpainting
Coherent Semantic Attention for image inpainting(ICCV 2019)
Stars: ✭ 202 (-11.4%)
Mutual labels: generative-adversarial-network
Transmomo.pytorch
This is the official PyTorch implementation of the CVPR 2020 paper "TransMoMo: Invariance-Driven Unsupervised Video Motion Retargeting".
Stars: ✭ 225 (-1.32%)
Mutual labels: generative-adversarial-network
Generative inpainting
DeepFill v1/v2 with Contextual Attention and Gated Convolution, CVPR 2018, and ICCV 2019 Oral
Stars: ✭ 2,659 (+1066.23%)
Mutual labels: generative-adversarial-network
Neuralnetworks.thought Experiments
Observations and notes to understand the workings of neural network models and other thought experiments using Tensorflow
Stars: ✭ 199 (-12.72%)
Mutual labels: generative-adversarial-network
Graph Representation Learning
Autoencoders for Link Prediction and Semi-Supervised Node Classification (DSAA 2018)
Stars: ✭ 199 (-12.72%)
Mutual labels: semi-supervised-learning
Arbitrary Text To Image Papers
A collection of arbitrary text to image papers with code (constantly updating)
Stars: ✭ 196 (-14.04%)
Mutual labels: generative-adversarial-network
Improved GAN (Semi-supervised GAN)
This is an implementation of Semi-supervised generative adversarial network in the paper Improved Techniques for Training GANs for Mnist dataset. This method and its extensions have marvellous performance on traditional CV datasets, and remain state-of-art (by the end of November, 2017).
Working Principle
Inspired by Good Semi-supervised Learning that Requires a Bad GAN, semi-supervised GAN with feature matching actually generates unrealistic fake samples around high-density region. With the inborn continuity, the fake region in feature space split the bounds of different classes.
Refer to Semi-supervised Learning on Graphs with Generative Adversarial Nets for more details about this density gap splitting explaination.
Running
The code was implemented in Python 3.7.
python ImprovedGAN.py
Default configs include CPU, saving and autoloading, generating logfile in tensorboard format, etc. You can use python ImprovedGAN.py --cuda to run it on GPU.
The latest torch(1.2 version), tensorboardX, torchvision are needed.
Result
Default configs can train models achieving 98.5% accuracy on test dataset with 100 labeled data(10 per class) and other 59,000 unlabeled data after 100 epochs.
Loss curve during training
loss_label => red, loss_unlabel => blue, loss_gen => green
It must be noted that OpenAI implementation(theano) demonstrates a different curve, where loss_gen is nearly zero and loss_unlabel increase gradually.
Remark
- The implementation is based on OpenAI implementation.
- But I found it hard to reproduce expected results and suffered from exploding gradients. I changed the final layer in generator from Sigmoid to Softplus, and therefore fixed it.
-
./modelsincludes the trained model, you can simply delete it for retraining. - The archectures of networks are elaborately designed, among them
Weight Normalizationis very important. - Thank Jiapeng Hong for discussing with me.
Change Logs
- (Nov 27, 2019) Update to pytorch 1.2 and Python 3.7. The version for pytorch 0.3 and Python 2.7 can be found in the history versions. Delete pretrained models.
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].