Autoencoders using PyTorch

Quick Links

• About
• Setup
  • Installation
  • Training
• Results

About

Fig 1: Architecture of an Autoencoder

Setup

Installation

1. Download the GitHub repo by using the following command running from the terminal.

git clone https://github.com/arpanmukherjee/Autoencoders-and-more-using-PyTorch.git
cd Autoencoders-and-more-using-PyTorch/

2. Install pip from the terminal, for more details please look here. Go to the following project folder and install all the dependencies by running the following command. By running this command, it will install all the dependencies you will require to run the project.

pip install -r requirements.txt

Training

The network can be trained using main.py script. Currently, it only accepts the following arguments with the allowed values. Please strictly follow the argument name and any of the values.

argument	accepted values	default value
epochs	integer	75
batch-size	integer	16
learning-rate	float	0.001
seed	int	1
data-path	data directory	../dataset/
dataset	MNIST or STL10 or CIFAR10	-
use_cuda	bool	False
network-type	FC or Conv	FC
weight-decay	float	1e-5
log-interval	int	50
save-model	bool	True

Arguments that have no default value, you must provide value to run the script.

python main.py --dataset STL10 --use-cuda True --network-type FC

If you think the model is taking too much time, you can consider using GPU. Set use_cuda argument as True.

Results

The following are the resultant images for my trained model. I have trained my model on the MNIST dataset for 75 epochs.

python main.py --dataset MNIST --use-cuda True

Trained Model	Actual Image	Decoded Image
MNIST
CIFAR10