| library_name: tf-keras | |
| tags: | |
| - ImageGeneration | |
| - GauGAN | |
| - GAN | |
| - spatially-adaptive normalization | |
| - Encoder | |
| - Segmentation-maps | |
| ## Model description | |
| In this, GauGAN architecture has been implemented for conditional image generation which was proposed in [Semantic Image Synthesis with Spatially-Adaptive Normalization](https://arxiv.org/abs/1903.07291). | |
| GauGAN uses a `Generative Adversarial Network (GAN)` to generate realistic images that are conditioned on cue images and segmentation maps. | |
| This repo contains the model for the notebook [**GauGAN for conditional image generation**](https://keras.io/examples/generative/gaugan/) | |
| Full credits go to [Soumik Rakshit](https://github.com/soumik12345) & [Sayak Paul](https://twitter.com/RisingSayak) | |
| ## Training and evaluation data | |
| Here, the [Facades dataset](https://cmp.felk.cvut.cz/~tylecr1/facade/) is used for training GauGAN model. Some custom layers that were added into the model are - SPADE (SPatially-Adaptive (DE) normalization), Residual block including SPADE & Gaussian sampler. Also, the GauGAN encoder consists of a few downsampling blocks. It outputs the mean and variance of a distribution as shown in this [image](https://i.imgur.com/JgAv1EW.png). | |
| ## Training procedure | |
| ### Training hyperparameters | |
| The following hyperparameters were used during training: | |
| | name | learning_rate | decay | rho | momentum | epsilon | centered | training_precision | | |
| |----|-------------|-----|---|--------|-------|--------|------------------| | |
| |RMSprop|0.0010000000474974513|0.0|0.8999999761581421|0.0|1e-07|False|float32| | |
| ## Model Plot | |
| <details> | |
| <summary>View Model Plot</summary> | |
|  | |
| </details> | |
| <center> | |
| Model Reproduced By <u><a href="https://github.com/robotjellyzone"><b>Kavya Bisht</b></a></u> | |
| </center> |