Tatiana Shpilevaya

The application of deep learning neural networks enables the processing of extensive volumes of data and often requires dense datasets. In certain domains, researchers encounter challenges related to the scarcity of training data, particularly in marine biology. In addition, many sounds produced by sea mammals are of interest in technical applications, e.g., underwater communication or sonar construction. Thus, generating synthetic biological sounds is an important task for understanding and studying the behavior of various animal species, especially large sea mammals, which demonstrate complex social behavior and can use hydrolocation to navigate underwater. This study is devoted to generating sperm whale vocalizations using a limited sperm whale click dataset. Our approach utilizes an augmentation technique predicated on the transformation of audio sample spectrograms, followed by the employment of the generative adversarial network StyleGAN2-ADA to generate new audio data. The results show that using the chosen augmentation method, namely mixing along the time axis, makes it possible to create fairly similar clicks of sperm whales with a maximum deviation of 2%. The generation of new clicks was reproduced on datasets using selected augmentation approaches with two neural networks: StyleGAN2-ADA and WaveGan. StyleGAN2-ADA, trained on an augmented dataset using the axis mixing approach, showed better results compared to WaveGAN.

Non-photorealistic rendering (NPR) with explicit brushstroke representation is essential for both high-grade imitating of artistic paintings and generating commands for artistically skilled robots. Some algorithms for this purpose have been recently developed based on simple heuristics, e.g., using an image gradient for driving brushstroke orientation. The notable drawback of such algorithms is the impossibility of automatic learning to reproduce an individual artist’s style. In contrast, popular neural style transfer (NST) algorithms are aimed at this goal by their design. The question arises: how good is the performance of neural style transfer methods in comparison with the heuristic approaches? To answer this question, we develop a novel method for experimentally quantifying brushstroke rendering algorithms. This method is based on correlation analysis applied to histograms of six brushstroke parameters: length, orientation, straightness, number of neighboring brushstrokes (NBS-NB), number of brushstrokes with similar orientations in the neighborhood (NBS-SO), and orientation standard deviation in the neighborhood (OSD-NB). This method numerically captures similarities and differences in the distributions of brushstroke parameters and allows comparison of two NPR algorithms. We perform an investigation of the brushstrokes generated by the heuristic algorithm and the NST algorithm. The results imply that while the neural style transfer and the heuristic algorithms give rather different parameter histograms, their capabilities of mimicking individual artistic manner are limited comparably. A direct comparison of NBS-NB histograms of brushstrokes generated by these algorithms and of brushstrokes extracted from a real painting confirms this finding.