CAN: Creative Adversarial Networks, Generating "Art" by Learning About Styles and Deviating from Style Norms

By Ahmed M. Elgammal, Bingchen Liu, Mohamed Elhoseiny, Marian Mazzone. ICCC, 2017

604 citations recorded by the source corpus.

We propose a new system for generating art. The system generates art by looking at art and learning about style; and becomes creative by increasing the arousal potential of the generated art by deviating from the learned styles. We build over Generative Adversarial Networks (GAN), which have shown the ability to learn to generate novel images simulating a given distribution. We argue that such networks are limited in their ability to generate creative products in their original design. We propose modifications to its objective to make it capable of generating creative art by maximizing deviation from established styles and minimizing deviation from art distribution. We conducted experiments to compare the response of human subjects to the generated art with their response to art created by artists. The results show that human subjects could not distinguish art generated by the proposed system from art generated by contemporary artists and shown in top art fairs. Human subjects even rated the generated images higher on various scales.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c2de603cdd1f9135edcca7e74bc5ced916047109

How researchers see this paper

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

In Cognitive Neuroscience

In Computer Vision and Pattern Recognition

Citations in the corpus

Generative Adversarial Networks–Enabled Human–Artificial Intelligence Collaborative Applications for Creative and Design Industries: A Systematic Review of Current Approaches and Trends (2021, positive): “The system builds on existing approaches to the generation of novelty by GANs (Elgammal et al., 2017) and facilitates a degree of ﬁ ne control over the output and the degree of novelty introduced.”
Deep Generative Models in Engineering Design: A Review (2021, background): “For example, Creative Adversarial Networks (CAN) [177] introduce entropy into the”
Self-Consuming Generative Models Go MAD (2023, background): “Pictured here are representative samples from LAION-5B that include (clockwise from upper left and highlighted in red) synthetic images from the generative models StyleGAN [1], AICAN [35], Pix2Pix [36], DALL-E [37], and BigGAN [38].”
Deep Learning Techniques for Music Generation - A Survey (2017, methodology): “[39], as an extension of the generative adversarial networks (GAN) architecture (introduced in Section 5.”
Understanding and Creating Art with AI: Review and Outlook (2021, methodology): “For example, the authors of the AICAN system performed a sort of visual Turing test to explore if people can tell the difference between AICAN- and human-made art [45].”
Copyright in generative deep learning (2021, neutral): “In this context, the importance of adversarial training fromGANs and searching for diverging from existingworks as done by Elgammal et al. (2017) may tip the balance toward legality. With respect to other classic techniques, the generative part of a GAN never uses protected data; therefore, it is ha”
Generative Art Using Neural Visual Grammars and Dual Encoders (2021, background): “An interesting variant to the GAN approach is the Creative Adversarial Network (CAN) which adds a cost for making images which are too easily classiﬁable as an existing artistic genre [7].”
Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis (2021, methodology): “We hope this work can beneﬁt downstream tasks of GAN (Liu et al., 2020c;b; Elgammal et al., 2017) and provide new study perspectives for future research. attention to all the regions of an input image, and encode the image with a minimum information lost, then the decoder is easier to reconstruct…”