Image Reconstruction using AutoEncoders

Table of Contents

1. Reconstruction of CIFAR10 using AutoEncoder
2. Reconstruction of input Images from the average of the input images (CIFAR10)

Image Reconstruction

The repo contains a ipynb file for simple image reconstruction of cifar 10 using pytorch. Simple MSE loss is used with a single decoder. Skip connections are also used. Evaluation is done using SSIM, L1 and PSNR.

Input reconstruction from average image

The repo contains a ipynb file for input image reconstruction of cifar 10 from the average of the inputs using pytorch. A weighted loss of SSIM and L1 is used. I have used 2 decoders to reconstruct the two inputs seperately. Skip connections are also used. Evaluation is done using SSIM, L1 and PSNR.