diff --git a/README.md b/README.md index 57884980d12716147ef767baaee44c3afe0e851a..dababe861157297742e734b82bc17bfadebb24fd 100644 --- a/README.md +++ b/README.md @@ -1,7 +1,71 @@ -# GAN & cGAN tutorial. +# GAN & cGAN tutorial +MSO 3.4 Machine Learning -We recommand to use the notebook (.ipynb) but the Python script (.py) is also provided if more convenient for you. -# How to submit your Work ? +[](https://rfarssi.mit-license.org/) +[](https://www.python.org/) -This work must be done individually. The expected output is a repository named gan-cgan on https://gitlab.ec-lyon.fr. It must contain your notebook (or python files) and a README.md file that explains briefly the successive steps of the project. The last commit is due before 11:59 pm on Wednesday, March 29, 2023. Subsequent commits will not be considered. \ No newline at end of file +# Description +The aim of this assignment is to discover GANs, understand how they are implemented, and then explore one specific architecture of GANs that allows us to perform image-to-image translation. The current assignment consists of two parts : + +- The first part focuses on understanding the fundamental concepts of Generative Adversarial Networks (GANs) through a DCGAN example. + +- In the second part, we will implement and train a conditional GAN (cGAN) to generate facades. + +# Part 1 : DC-GAN +In this section, we will dive deeper into Generative Adversarial Networks and explore how they can be used to generate new images. We will be using a DCGAN (Deep Convolutional Generative Adversarial Network) to generate new images of handwritten digits using the [MNIST](https://pytorch.org/vision/stable/generated/torchvision.datasets.MNIST.html) dataset. + +To complete this task, we will need to retrain the DCGAN and generate some samples of automatically generated handwritten digits. We did this by following the PyTorch [DCGAN tutorial](https://pytorch.org/tutorials/beginner/dcgan_faces_tutorial.html). + +## Losses during training of the generator and discriminator + + + +## Visual Comparison between Real and Generated Images + + + + + +# Part 2 : Conditional GAN (cGAN) +This section involves training a cGAN to distinguish between real and generated building facades images from the [CMP Facade Database](https://cmp.felk.cvut.cz/~tylecr1/facade/). The generator takes a mask from the database and produces a realistic image as output. Here are some of the images obtained as a result of this process. + +- **100 epochs** + + + + +- **200 epochs** + + + +With a larger number of epochs applied to the training set, the quality of the generated images is expected to improve. + + + + + + + + + + + +# About the Project + +## Project status +[](https://gitlab.ec-lyon.fr/rfarssi/image-classification) + +## Contribution +Any contributions that improve the quality of this project are welcome [here](https://gitlab.ec-lyon.fr/rfarssi/mso3_4-be2_cgan/-/issues). + +## References +- [GAN & cGAN tutorial](https://gitlab.ec-lyon.fr/edelland/mso3_4-be2_cgan) by _Emmanuel Dellandrea_ + +## License +[](https://rfarssi.mit-license.org/) was used to grant permission for this project. + +## Author +[© Rahma FARSSI](https://gitlab.ec-lyon.fr/rfarssi) + +<a href="mailto:rahma.farssi@master.ec-lyon.fr? &body=Hi Rahma">📧</a>