| YOHANSSEN PRATAMA | D, 中間発表 | 光メディアインタフェース | 向川 康博, 清川 清, 舩冨 卓哉, 藤村 友貴, 北野 和哉 |
|
title: Time Series Deep Learning using Single Photon Transient Histogram for Per-Pixel Material Classification
abstract: In this research, we present a novel method of material classification through the transient histogram captured by a Single-Photon Avalanche Diode (SPAD) camera. By utilizing a single photon avalanche diode (SPAD), we can observe reflected light with a high time resolution of approximately 10 picoseconds which cannot be measured using a Time of Flight (ToF) camera. Due to the complex and nonlinear nature of the transient histogram shapes, traditional image analysis methods may not be suitable for this approach. Our strategy here is by utilizing the distinct transient signatures of each material to acquire some feature vector in the picosecond time range. After we get the transient histogram dataset for each material then the per-pixel material classification can be conducted by using one-dimensional deep-learning models for time series classification. language of the presentation: English | |||
| LUGTENBERG GEERT | D, 中間発表 | インタラクティブメディア設計学 | 加藤 博一, 清川 清, 神原 誠之, 藤本 雄一郎, 澤邊 太志 |
|
title: The effects of parallax depth cues on AR supported hand tasks using a non-wearable display
abstract: Non-wearable displays (i.e. fixed displays) have ergonomic, commercial and environmental benefits over popular head-worn displays (HMD) in its use of augmented imagery. However, non-wearable displays consist of mostly video see-through systems which in turn suffer from a wide range of challenges when used for hand task support. One such challenge is the incongruity of the displayed image and the surrounding context and its effect on depth perception. We argue that using a system in which the displayed image is from the perspective of the user's eye, instead of a static perspective, will increase depth awareness and accuracy in hand-related touch tasks, by providing a parallax depth cue that the conventional magic lens system does not have. Furthermore we expect multimodal interaction between visual and proprioceptive depth cues to provide a user with a cognitive map of the environment. We therefore evaluate the influence of the rendering perspective condition on the speed of acquiring such a cognitive map. language of the presentation: English | |||
| PORNMANEERATTANATRI SORATOUCH | D, 中間発表 | ソフトウェア設計学 | 飯田 元, 藤川 和利, 市川 昊平, 高橋 慧智 |
|
title: Automatic code modifications to improve program execution speed and efficiency
abstract: Parallel programming is essential to utilize multi-core processors but remains challenging because it requires extensive knowledge of both software and hardware. Various automatic parallelization tools based on static analysis have been developed to ease the development of parallel programs. However, hand-parallelized codes still outperform auto-parallelized codes. Meanwhile, transformer-based large language models have made ground-breaking progress in coder understanding and generation tasks. In this research, we fine-tune a transformer-based code understanding model, CodeT5, to create a model for automatically identifying parallelizable for-loops. The trained model helps developers to identify independent for-loops that can be potentially parallelized and generate parallelization directive using tools such as OpenMP to improve the program performance. language of the presentation: English | |||