| 日時(Datetime) | 令和4年6月23日(木)3限 (13:30 -- 15:00), 2022/06/23, Thursday |
|---|---|
| 場所(Location) | エーアイ大講義室(L1) |
| 司会(Chair) | 日永田 智絵(Chie Hieida) |
| 講演者(Presenter) | 嶋利 一真(Kazumasa Shimari), ソフトウェア工学研究室 (Software Engineering Lab.) |
|---|---|
| 題目(Title) | Near-Omniscient Debugging Using Size-Limited Execution Trace |
| 概要(Abstract) | In recent years, software development costs have increased, and the debugging costs, which account for half of software development costs, have also increased accordingly. Then, many debugging techniques have been proposed to reduce debugging costs. In this lecture, I introduce typical debugging techniques and our proposed debugging technique. We proposed Near-Omniscient Debugging which monitored a Java program's execution within limited storage space. We confirmed the usefulness of Near-Omniscient Debugging by measuring the data dependencies' accuracy of the execution traces, the amount of execution traces, and the percentage of bugs whose bug-related instructions were completely recorded. Using this debugging method, we have implemented a tool named NOD4J which annotates the source code with observed values in an HTML format. We show two examples that our tool can debug defects using incomplete execution traces. |
| 講演者(Presenter) | 北野 和哉(Kazuya Kitano), 光メディアインタフェース研究室 (Optical Media Interface Lab.) |
|---|---|
| 題目(Title) | Generalization of Super-resolution for Multimodal Optical Measurement System |
| 概要(Abstract) | Various optical phenomena are indicate the physical properties of an object. Hence, analytical methods that are based on observing the temporal and spectral responses of light are widely employed in the field of science. In general, a method for observing the optical response involves the usage of image sensors. However, the image sensors are near the limit of improvement due to the physical limitaiton. In this study, we propose a generalized super-resolution method for multimodal optical measurement systems. The proposed method was designed by generalizing the underlying principle of image super-resolution. We implemented the measurement systems with respect to the temporal and spectral axes for demonstrating the effectiveness of the proposed method. We designed a sub-sampling method to enable indirect gate shifting in the temporal axis and implemented the same for a time-of-flight camera. The obtained results indicate that we can achieve a super-resolution of 0.25 ns from the 20 ns sampling width in real environments. With respect to the spectral axis, we tilted the area sensor for realizing accurate sub-pixel shifting and for recovering a high-resolution spectrum by employing interpolated spectrally-varying kernels. Based on experimental validation, it is found that the proposed method can achieve a high spectral resolution of 12.9x super-resolution in the 400 nm–800 nm range through the mere tilting of the sensor in the spectrometer. |