This repository contains Master Projectwork ⬇️ undertaken for the partial fulfillment of M.S. in Information Engineering @Fachhochschule-Kiel under the supervision of Prof. Dr. Hauke Schramm
Objective : Series of progressive exploration and experimentation of Deep Generative Models on subset of Wiki-Art dataset to produce Realistic art Images.
Source : Wiki-Art : Visual Art Encyclopedia
Dataset Contents for Unconditional Generation
'abstract': 14999,
'animal-painting': 1798,
'cityscape': 6598,
'figurative': 4500,
'flower-painting': 1800,
'genre-painting': 14997,
'landscape': 15000,
'marina': 1800,
'mythological-painting': 2099,
'nude-painting-nu': 3000,
'portrait': 14999,
'religious-painting': 8400,
'still-life': 2996,
'symbolic-painting': 2999
Resized subsets for Conditional Generation
abstract
,cityscape
,landscape
,portrait
( no. of samples > 5000 )
photo2fourcollection for multi-collection style transfer
train_content
,test_content
,train_styles
-
-
FullyConnected MLP-GAN
-
DeepConvolutional-GAN
-
Wasserstein-GAN
- Weight Clipping
- Gradient Penalty
-
-
- Conditional-DCGAN
-
- Neural Style Transfer
- Multi Collection Style Transfer( GatedGAN )
-
[ 1 ] Generative Adversarial Nets [ arXiv:1406.2661v1 [stat.ML] 10 Jun 2014 ]
-
[ 2 ] Conditional Generative Adversarial Nets [arXiv:1411.1784v1 [cs.LG] 6 Nov 2014]
-
[ 4 ] Improved Techniques for Training GANs [ arXiv:1606.03498v1 [cs.LG] 10 Jun 2016 ]
-
[ 5 ] Wasserstein GAN [ arXiv:1701.07875v3 [stat.ML] 6 Dec 2017 ]
-
[ 6 ] Improved Training with WGAN-GP [ arXiv:1704.00028v3 [cs.LG] 25 Dec 2017 ]
-
[ 8 ] VGG-19 as Feature(style/semantics) Extractor - PyTorch Documentation
-
[ 9 ] Coursera (Deeplearning.ai) - Build Basic GANs [ Online Course ]
-
[ 11 ] GitHub - HarisIqbal88/PlotNeuralNet: Latex code for making neural networks diagrams