Megapixel Size Image Creation using Generative Adversarial Networks

By Marco Marchesi. 2017

85 citations recorded by the source corpus.

Since its appearance, Generative Adversarial Networks (GANs) have received a lot of interest in the AI community. In image generation several projects showed how GANs are able to generate photorealistic images but the results so far did not look adequate for the quality standard of visual media production industry. We present an optimized image generation process based on a Deep Convolutional Generative Adversarial Networks (DCGANs), in order to create photorealistic high-resolution images (up to 1024x1024 pixels). Furthermore, the system was fed with a limited dataset of images, less than two thousand images. All these results give more clue about future exploitation of GANs in Computer Graphics and Visual Effects.

Subfields: Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

In Generative Models

Citations in the corpus

Training Generative Adversarial Networks with Limited Data (2020, background): “We can see that ADA yields excellent results in all cases, and with slight truncation [29, 20], virtually all of the images look convincing.”
A Style-Based Generator Architecture for Generative Adversarial Networks (2018, methodology): “Here we used a variation of the truncation trick [38, 4, 31] with ψ = 0.”
Large Scale GAN Training for High Fidelity Natural Image Synthesis (2018, background): “Related observations about this trade-off were made in (Marchesi, 2016; Pieters & Wiering, 2014).”
Progressive Growing of GANs for Improved Quality, Stability, and Variation (2017, methodology): “enerated using the CELEBA-HQ dataset. See Appendix F for a larger set of results, and the accompanying video for latent space interpolations. On the right, two images from an earlier megapixel GAN by Marchesi (2017) show limited detail and variation. Mao et al. (2016b) (128 128) Gulrajani et al. (20”
Improved Precision and Recall Metric for Assessing Generative Models (2019, methodology): “For instance, it is interesting that while recent state-of-the-art generative methods [4, 14, 13] claim to optimize FID, in the end the (uncurated) results are almost always produced using another model that explicitly sacrifices variation, and often FID, in favor of higher quality samples from a tr”
The Effectiveness of Data Augmentation in Image Classification using Deep Learning (2017, background): “Additionally, they have shown to be extremely good at augmenting data sets, such as increasing the resolution of input images [13].”
Scenic: a language for scenario specification and scene generation (2018, background): “Second, we pick a position for the bottleneck, requiring it to lie roughly on the way from the robot to its goal, and place a rock there.”
StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis (2023, methodology): “Typically variation has been traded to higher ﬁdelity in GANs using the truncation trick (Marchesi, 2017; Brock et al., 2019; Karras et al., 2019), where a sampled latent w is interpolated towards its mean with respect to the given conditioning input.”
Efficient Geometry-aware 3D Generative Adversarial Networks (2021, methodology): “By adding generator pose conditioning (third row), we allow the generator to faithfully model pose-correlated attributes while decoupling them at inference, leading to both the best FID score and viewconsistent results.”
Counterexample-Guided Data Augmentation (2018, methodology): “More recently, a sophisticated data augmentation technique has been proposed [Liang et al. , 2017; Marchesi, 2017] which uses Generative Adversarial Networks [Goodfellow et al. , 2014], a particular kind of neural network able to generate synthetic data, to inﬂate training sets.”
Instance Selection for GANs (2020, background): “Perhaps the most well known approach for increasing sample ﬁdelity in GANs is the “truncation trick” [2, 10, 15].”