Neural Policy Style Transfer

Fernandez-Fernandez, Raul; Victores, Juan G.; Gago, Jennifer J.; Estevez, David; Balaguer, Carlos

doi:10.1016/j.cogsys.2021.11.003

Cited by 5 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhang et al improved AdaIN in style control by fixing the image optimisation target and setting cross-channel correlation, and comprehensively implemented Adaptive Style Modulation (AdaSM) [15] to enhance global style control, but large numbers of artefacts appeared in the field of mural applications. Fernandez et al introduced style migration into the field of reinforcement learning, and utilised the Neural Policy Style Migration algorithm (NPST) [16] to achieve style migration to the field of intelligent robotics, but the model is not applicable to static mural effects. Yu et al added the Covariance Attention Network (CovAttN) [17] to achieve stylistic uniformity between different patches of the image in terms of stylistic fusion, but stylistic confusion occurs when facing mural paintings where styles may be misaligned, which affects the final generation effect.…”

Section: Image Style Migrationmentioning

confidence: 99%

Feature Separation and Fusion to Optimise the Migration Model of Mural Painting Style in Tombs

Wu,

Li,

Zhang

2024

Applied Sciences

View full text Add to dashboard Cite

Tomb murals are different from cave temple murals and temple murals, as they are underground cultural relics, their painting style is unique, solemn, and austere, and the performance image is characterised by simple colours, low contrast, and fewer survivors. During the digital restoration process, it is important to have sufficient reference samples to ensure the accuracy of the restoration. In addition, the style of mural paintings in the tombs varies greatly from other styles of murals and types of word paintings. Therefore, learning the unique artistic style of tomb murals, providing stylistically consistent training samples for digital restoration, and overcoming the problems of dim lighting and complex surface granularity of tomb murals are all necessary for research. This paper proposes a generative adversarial network algorithm that separates and fuses style features to enhance the generative network’s ability to acquire image information. The algorithm extracts underlying and surface style feature details of the image to be tested and conducts fusion generation experiments. The generative network’s parsing layer modifies the input noise tensor and optimises the corresponding weights to prevent misalignment between drawing lines and fresco cracks. Finally, to optimise the fresco generation effect, we add the corresponding loss function in the discriminator. The tomb murals dataset was established for experiments and tests, and quantitatively and qualitatively analysed with other style migration models, and SSIM, FID, LPIPS and NIQE were used as evaluation indexes. The results were 0.97, 269.579, 0.425 and 3.250, respectively, and the effect of style migration of this paper’s method was significantly higher than that of the control group model.

show abstract

Section: Image Style Migrationmentioning

confidence: 99%

Feature Separation and Fusion to Optimise the Migration Model of Mural Painting Style in Tombs

Wu,

Li,

Zhang

2024

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…The Neural Policy Style Transfer (NPST)1 algorithm is capable of transferring style while preserving content. Experiments are conducted in catchball games and real-life situations (Fernandez-fernandez et al, 2022). Utilizing Neural Style Transferring algorithms, the study proposes a method for generating fully aligned synthetic multispectral images from authentic misaligned unmanned aerial vehicle images.…”

Section: Literature Reviewmentioning

confidence: 99%

Enhancing Supervised Learning through Empirical Enrichment Using Style Transfer Generative Datasets

Hindarto

2024

SinkrOn

View full text Add to dashboard Cite

An innovative strategy for improving supervised learning by utilizing empirically enriched datasets through the application of generative style transfer techniques. Within the realm of artificial intelligence, supervised learning has emerged as a significant domain. However, the challenge of acquiring datasets that are both representative and diverse persists. To tackle this issue, this research integrates the notion of style transfer to broaden the range of data accessible for supervised learning models. This method employs the style transfer process to generate diverse style variations within the existing data. Incorporating various image variations enhances the dataset and enables the model to gain a deeper comprehension of the image's content. Experiments were performed utilizing a conventional dataset that was enhanced using a style transfer technique and subsequently inputted into a supervised learning model. The results demonstrate substantial enhancements in model performance, particularly in terms of its ability to generalize to new test data. This confirms the efficacy of this approach in enhancing the quality of supervised learning. These findings emphasize the significant potential of employing style transfer in dataset enrichment to improve and intensify model comprehension in managed learning scenarios, as well as its implications in the advancement of artificial intelligence technologies that are more flexible and capable of adjusting to various visual scenarios.

show abstract

“…However, these studies are product-specific and lack a general methodology for designing motions across a broad range of products. In robotics research, the generation of robot arm motions has been explored to express emotions and enhance human interaction (Fernandez et al 2023;Hagane and Venture 2022;Li and Zhao 2023), and in human-computer interaction (HCI), emotions estimated from speech have been used to guide avatar gestures (Qi et al, 2023). While these approaches detail methods for integrating emotional expressions into motion, they primarily aim to convey predefined emotions from robots to humans rather than inducing specific emotions in users through motion design.…”

Section: Introductionmentioning

confidence: 99%

Inquisitive Motion Generation System based on a Free-Energy Model of Curiosity and Interest

Honda,

Yanagisawa

2024

Preprint

View full text Add to dashboard Cite

Product motion is a crucial design element, blending functionality with aesthetic appeal. Such motions establish an emotional bond between the product and its user. Engaging with the product stimulates curiosity and interest in its motion, encouraging repeated use. This study aims to formulate indicators for curiosity and interest in motion and to implement a computational system to generate motions that provoke inquiry. This study situates curiosity and interest as emotions crucial to the inquiry cycle, proposing that optimal prediction error (deviation from expected motion patterns) cultivates interest, while fluctuations in prediction error spark curiosity. Our analysis is based on the inquiry cycle, a mathematical model of emotions rooted in the free energy principle of brain function. We developed curiosity, motion, and interest measures from the model based on prediction error variation and reduction. The developed system can independently control prediction errors and their variations for a given open-loop mechanism. This system generates motions that control the magnitude of prediction errors by deviating from the trajectory of energy-minimizing motions. Using the semantic differential method and eye-tracking experiments measuring gaze duration, we found that interest correlates with prediction error in an upward-convex manner, and its variability induces curiosity. These results align with our model. The novelty of this study lies in experimentally verifying that applying moderate prediction errors and fluctuations in prediction errors is effective in increasing interest and curiosity in motion design. This research contributes design insights for engineers and designers seeking to enhance the appeal of product motions.

show abstract

Neural Policy Style Transfer

Cited by 5 publications

References 11 publications

Feature Separation and Fusion to Optimise the Migration Model of Mural Painting Style in Tombs

Feature Separation and Fusion to Optimise the Migration Model of Mural Painting Style in Tombs

Enhancing Supervised Learning through Empirical Enrichment Using Style Transfer Generative Datasets

Inquisitive Motion Generation System based on a Free-Energy Model of Curiosity and Interest

Contact Info

Product

Resources

About