/AutoEncoder_MNIST_Fashion

Autoencoder using TensorFlow

Primary LanguageJupyter NotebookMIT LicenseMIT

AutoEncoder_MNIST_Fashion

A basic implementation of an Autoenconder to visualize the concept. In the past decade Image retrieval is the advancing Field of research in the computer vision domain. Apart from the CNNs, the Autoencoders plays a significant role for such retrieval systems. Autoencoder is a Neural Network and also an unsupervised Learning Algorithm. It applies Backpropogation algorithm by settiing the same input as the target value.

Installation and running

Install Tensorflow for GPU machine

pip install tensorflow-gpu

If you don't have a GPU

pip install tensorflow

Download the MNIST Fashion Dataset https://www.kaggle.com/zalando-research/fashionmnist

Put the proper path of the dataset in the autoencoder.py file , and type

python autoencoder.py

Now, training will start, and at the end you can observe the output generated by the decoder, which will be similar to the input image provided.

Once you have trained it, you can visualize the output of the decoder and compare it with the original image. Here in the figure below, left image is the original, and one in the right is created by reshaping the output vector of size 784 into 28x28. Well! now one can use the encoder part as a pretrained network.

Decoder_ouput