It is a simple application that utilizes Convolutional AutoEncoder for extracting latent features by doing unsupervised training on CIFAR-10 dataset. After the Images are transformed using the Encoder part of ConvAutoEncoder into 512-D vector, ANN is used to find similar images as the given one. The Code for extracting features is inside CIFAR_CAE.ipynb that will save the trained model as ConvAutoEncoder.h5. Code for similar image-search using ANN is inside ImageSearch. Pre-trained is provided in the repo, choose to use that or else train your own model by changing hyperparameters.
