Evolutionary Latent Space Exploration of Generative Adversarial Networks

Fernandes, Paulo; Correia, João; Machado, Penousal

doi:10.1007/978-3-030-43722-0_38

Cited by 18 publications

(8 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous work has applied genetic algorithms to explore the latent space with different goals. While Roziere et al [24] improve the quality of only one fashion image generated by a P-GAN, Fernandes et al [7] evolve a set of latent vectors to increase the diversity among the corresponding set of generated images. Instead of a pre-defined fitness objective, interactive genetic algorithms use user evaluation as a measure of fitness.…”

Section: Evolutionary Search Of Gans' Latent Spacementioning

confidence: 99%

See 1 more Smart Citation

Fashion Style Generation: Evolutionary Search with Gaussian Mixture Models in the Latent Space

Grabe¹,

Zhu²,

Agirrezabal³

2022

Preprint

View full text Add to dashboard Cite

This paper presents a novel approach for guiding a Generative Adversarial Network trained on the FashionGen dataset to generate designs corresponding to target fashion styles. Finding the latent vectors in the generator's latent space that correspond to a style is approached as an evolutionary search problem. A Gaussian mixture model is applied to identify fashion styles based on the higher-layer representations of outfits in a clothing-specific attribute prediction model. Over generations, a genetic algorithm optimizes a population of designs to increase their probability of belonging to one of the Gaussian mixture components or styles. Showing that the developed system can generate images of maximum fitness visually resembling certain styles, our approach provides a promising direction to guide the search for style-coherent designs.

show abstract

Section: Evolutionary Search Of Gans' Latent Spacementioning

confidence: 99%

“…While the following sections explain the details, Algorithm 1 shows the pseudo-code for the procedure. 7 Representation and transformation A generated design is represented by its latent vector z = v 1 , ..., v l with v ∈ R consisting of l latent variables.…”

Section: Evolutionary Searchmentioning

confidence: 99%

Fashion Style Generation: Evolutionary Search with Gaussian Mixture Models in the Latent Space

Grabe¹,

Zhu²,

Agirrezabal³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This can be done by operating directly on the latent codes [15,16] or by analysing the activation space of latent codes to discover interpretable directions of manipulation in latent space [17]. Evolutionary methods have been applied to search and map the latent space [18,19] and interactive evolutionary interfaces have also been built to operate on the latent codes [20] for human users to explore and generate samples from generative models.…”

Section: Related Work 21 Deep Generative Modelsmentioning

confidence: 99%

Network Bending: Expressive Manipulation of Deep Generative Models

Broad

Leymarie

Grierson

2021

Artificial Intelligence in Music, Sound, Art and Design

View full text Add to dashboard Cite

We introduce a new framework for manipulating and interacting with deep generative models that we call network bending. We present a comprehensive set of deterministic transformations that can be inserted as distinct layers into the computational graph of a trained generative neural network and applied during inference. In addition, we present a novel algorithm for analysing the deep generative model and clustering features based on their spatial activation maps. This allows features to be grouped together based on spatial similarity in an unsupervised fashion. This results in the meaningful manipulation of sets of features that correspond to the generation of a broad array of semantically significant features of the generated images. We outline this framework, demonstrating our results on state-of-the-art deep generative models trained on several image datasets. We show how it allows for the direct manipulation of semantically meaningful aspects of the generative process as well as allowing for a broad range of expressive outcomes.

show abstract

“…This can be done by operating directly on the latent codes [ 17 , 18 ] or by analysing the activation space of latent codes to discover interpretable directions of manipulation in latent space [ 19 ]. Evolutionary methods have been applied to search and map the latent space [ 20 , 21 ] and interactive evolutionary interfaces have also been built to operate on the latent codes [ 22 ] for human users to explore and generate samples from generative models.…”

Section: Related Workmentioning

confidence: 99%

Network Bending: Expressive Manipulation of Generative Models in Multiple Domains

Broad

Leymarie

Grierson

2021

Entropy

View full text Add to dashboard Cite

This paper presents the network bending framework, a new approach for manipulating and interacting with deep generative models. We present a comprehensive set of deterministic transformations that can be inserted as distinct layers into the computational graph of a trained generative neural network and applied during inference. In addition, we present a novel algorithm for analysing the deep generative model and clustering features based on their spatial activation maps. This allows features to be grouped together based on spatial similarity in an unsupervised fashion. This results in the meaningful manipulation of sets of features that correspond to the generation of a broad array of semantically significant features of the generated results. We outline this framework, demonstrating our results on deep generative models for both image and audio domains. We show how it allows for the direct manipulation of semantically meaningful aspects of the generative process as well as allowing for a broad range of expressive outcomes.

show abstract

Evolutionary Latent Space Exploration of Generative Adversarial Networks

Cited by 18 publications

References 7 publications

Fashion Style Generation: Evolutionary Search with Gaussian Mixture Models in the Latent Space

Fashion Style Generation: Evolutionary Search with Gaussian Mixture Models in the Latent Space

Network Bending: Expressive Manipulation of Deep Generative Models

Network Bending: Expressive Manipulation of Generative Models in Multiple Domains

Contact Info

Product

Resources

About