Utilizing deep learning in sports match analysis is very promising for efficiency, and there have been many studies focusing on volleyball. However, previous studies have used only the frames where the ball is alive (i.e., in-play frames), and recent data analysis methodologies rely on visual data, which are highly dependent on the position and angle of the camera relative to the court. As a result, these methods require a large dataset of images taken from various angles to improve accuracy. To efficiently extract and analyze the play as it is happening, we propose a model that distinguishes between in-play and out-of-play by combining visual data and audio data of volleyball match videos using late fusion. To investigate the effectiveness of the proposed model, we perform Grad-CAM visualization to determine which pixels the proposed model is focusing on.

This paper introduces a multi-sized particle sampling method within an arbitrary 2D shape using power tessellation. Our method aims to improve packing density as to sample as many particles as possible in a limited area. We propose a porosity-driven optimization technique to ensure no overlap between particles while increasing the packing density. With such properties, our method is applicable to physically-based simulations, such as the Discrete Element Method and its related framework. Additionally, this technology allows users to set the target particle size distribution by a pre-designed cumulative distribution function and restrict the errors between 10% and 20% after the optimization. We demonstrate that our multi-sized particle sampling algorithms significantly improve packing density compared to Poisson disk sampling and SPH-based blue noise sampling.

Thin-film interference is widely known for the reason of the iridescence on a bubble and spilled oil. Although it is not difficult to calculate the interference of light waves, calculating high precision interference is still considered impractical in real-time rendering. The proposed method is to separate the interference calculation by pre-computing calculation and save the value in lookup texture. But instead of using angle and thickness like other methods, we are using optical path difference as an index in lookup texture to create a flexible and compact lookup texture. The rest of the calculation, which calculates the optical path difference from the thin-film attribute will be done in real-time. Compared with other methods, the lookup texture generated using optical path difference will consume significantly less memory since it reduces the texture dimension to a one-dimensional texture. With this, we can handle different parameter changes such as refractive index and reflectance value without generating a new lookup texture or impacting the rendering performance.

“Eyeliner,” a makeup method makes the eyes look larger, and makes the face so-called ‘attractive.’ Eyelids shapes of Asian women are roughly categorized into “double,” “hooded,” and “single.” We assumed that the most attractive eyeliner for her would depend on her eyelids shape. In this paper, we quantitatively measured the length and the thickness of eyeliner (eyeliner style) that make the eyes appear the largest without any sense of incongruity for each eyelids shape and investigated whether the eyeliner style is attractive. The purpose of this paper is to clarify attractive eyeliner styles by eyelids shape. The measurement was conducted using digital makeup with eyeliner models devised in this study. The results showed that the eyeliner that makes the eyes look the largest without any sense of incongruity is the most attractive eyeliner for each eyelids shape.