Skip to content

Latest commit

 

History

History
25 lines (18 loc) · 1009 Bytes

README.md

File metadata and controls

25 lines (18 loc) · 1009 Bytes

Semantic Inpainting using DCGANs


This is an experimental tensorflow implementation of semantic inpainting from corrupted images using DCGANs from the paper Semantic Image Inpainting with Perceptual and Contextual Losses. A major help was Brandon Amos blog on Image Completion.One of the major difference between is the training method used.I have used Adam Optimizer instead of gradient descent.


Requirements

  • Tensorflow
  • glob
  • Python 3

Dataset

  • I have used Celebrity faces dataset CelebA.Download the aligned version ,extract in the same directory as the code

Model Architecture

alt-text


Few Results are Partial Training

Note: Due to unavailabity of GPU,i didnt train the model for long.This results are after an hour of training

alt-text alt-text