Chris Pollett > Students >
Deshmukh

    ( Print View)

    [Bio]

    [Blog]

    [CS 297 Proposal]

    [Deliverable 1]

    [Deliverable 2]

    [Deliverable 3]

    [Deliverable 4]

    [End to End Image Compression - PDF]

    [Image file formats - PDF]

    [DeepN-JPEG - PDF]

    [Image compression using RNN - PDF]

    [CS 297 Report - PDF]

    [CS 298 Proposal]

    [CS 298 report]

















Deliverable 2: Implement Singele Image Super Resolution using GAN

Description: In this deliverable a single image super resolution algorithm proposed by C. Ledig [1] was implemented. We used the ready-made VGG network provided by pytorchvision module for discriminator network. As with any Generative Adversarial Network, the generator network of this framework is crucial for the task of super resolution.

Our architecture was implemented using pytorch framework and executed using google colab notebook as well as Google Cloud Platform. Images from the STL10 dataset were used for training as well as validation. A subset of images which were not used in training were used for inference purpose.

As the name suggests, class "generator" and "discriminator" are responsible for generator and discriminator respectively. "Residual_block" and "unsampled_block" classes are coded as per their definition in the research paper. Batch size was clapped to 16 due to limited resources.


The inference results obtained are as follows: :

High resolution fake images High resolution real images low resolution images
     
     
     

References :

  1. Ledig Christian, Et al. (May 2017) .Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network. Conference on Computer Vision and Pattern Recognition . Honolulu, Hawaii. arxiv e-archive