diff --git a/README.md b/README.md index b4233387eb8969cc8f56388d776a892c891f548e..e114e009b0d5369158002cf71dfd513eda9a7a63 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,18 @@ # gan-cgan This repository contains the work corresponding to the BE2 of the course MSO 3.4 - Deep structured Learning, by Julie Guerra. +It contains the Python notebook with the code and the answers to the theoritical questions, and two images showing results. + +The main results are below, in case it's impossible to rerun the Python notebook. + +## First part: DC-GAN + +The process for this assignment started with working on a DC-GAM with a database of handwritten numbers. Here is the result for 5 epochs of training: + + + +## Second part: cGAN + +The second part consisted in working on a Conditional GAN (cGAN) for the construction of the image of a building from a template. I built the cGAN and trained it for 200 epochs, saving the model for 100 and 200 epochs. Here is the comparison of the result of the network for 100 and 200 epochs, on an image from the validation dataset: + +