This paper aims to improve a teaching quality of a beginner teacher, quantitatively evaluates the difference between beginner teachers and specialist teachers from analyzing behaviors of beginner teachers and specialist teachers. Specifically, this study extracts behavior patterns of two groups using spatial-temporal graph convolution networks, and generates pattern datasets of two groups. As an experiment, we evaluated the teacher’s behavior by calculating the score of the trial lesson based on pattern datasets. As a result, we obtained a classification accuracy of 81.1 percents against beginner and specialist teacher data, were able to analyze teacher’s behavior. Furthermore, beginner teachers’ class consisted of the movement of “writing on the board”, on the other hand, specialist teachers’ class composed the class by combining “standing and speaking” with other behaviors. This visualization shows that beginner teachers can obtain awareness of the differences between experienced teachers and classes.

We propose a method for estimating camera trajectories optimally aligned with prepared environment maps and its application. We implement Direct Sparse Odometry (DSO) as real-time Visual Odometry (VO) and Structure from Motion (SfM) to build environment maps. In this process, a pose graph is created and optimized using related camera poses to calculate the alignment of all frames from DSO towards the environment map with the scales matched. Therefore, the proposed method runs solely with a monocular color camera and does not require any training datasets. We demonstrate that the proposed method is able to be used as an object-based recognition system based on the difference between a map as a prior information and the current scene observations by the user. We tested this real-time application in a laboratory environment. In addition, we conducted an experiment to evaluate the accuracy of the proposed method using an existing dataset. The results showed that the proposed method improved the accuracy of pose estimation compared to the case using only DSO.

Moiré is an interference pattern caused by the superposition of multiple periodic patterns. Although it is generally treated as an object to be removed by printing, it has been used in architectural design and art because of its visual interest. Recently, several methods have been proposed to apply moiré as an animation technique. However, existing moiré animation methods can only represent linear and beating motions, which limits the motion of the moiré. In this study, we propose a generalized method of moiré animation, which realizes motion along an arbitrary curved path and counterintuitive motion where the moiré moves back and forth despite the fact that the grating is moving only in one direction. In this study, we focus on moiré animation, which is generated by superposition of a moving grating composed by defining a moving waveform and repeating it regularly, and a fixed grating defined by a fixed waveform, phase shift, and intensity. A fixed waveform can be designed freely, and the brightness of moiré appears to move by continuously changing the phase shift in the region where moiré occurs. The proposed method allows for a wider range of expression than existing methods. Therefore, it is expected to be applied to eye-catching architectural designs and advertising applications.

The widespread of VR and hand tracking technology opens new possibilities of 3D modeling tools that enable a novice user to create a model with a flat learning curve. However, existing VR modeling tools are difficult for beginner to start modeling because of its complexity on trial-and-error based operations, such as adjustment of deformation on the model. In this paper, we propose a novel 3D modeling system that enables a user to create a freeform 3D shape based on a play-dough metaphor. The user can deform the shape of an object by simple two-handed hand gestures. The object is represented as a set of playdough particles and their connectivity changes according to the hand gestures. We show several examples of generated 3D models to demonstrate the expressiveness of our system. We also report the result of our pilot user study conducted to examine the acceptability to novice users.

In this paper, we propose a novel color-suggestion method considering the semantics of graphic design to support the starting point in color design. Although color design is important in design production, the standard OS color palette provided by paint applications has too many choices, which often leads to confusion in selection. Therefore, we propose a machine learning-based color generation model that can generate multiple color candidatesusing the layer names set for each layer of a given graphic design. Our method presents candidate colors to the users by considering the semantics of the graphic design for each layer. Therefore, the users can obtain the starting color of the graphic design exploration, and can obtain diverse designs. We implement a web application using our method and evaluate the extent to which the method supports the design process of novice users.