My purpose is to combine cutting edge technologies developed by researches, and create accessible packages so that anyone who might not be interested in the science can also benefit and appreciate the technologies.
Experience:
NISSAN MOTOR CORPORATION internship | Invisible-to-Visible Project (I2V) | AR & NLP application development
Development of a Real-Time Control Software for THz Radar that takes the full potential of its hardware performance.
45th International Conference on Infrared, Millimeter, and Terahertz Waves 2020 | oral presenter
35th conference of the society of instrument and control engineers | oral presenter
LinkedIn Profile View ResumeDeveloped a real-time control software for a novel terahertz radar system using multi-processing to handle data processing and visualization without losing the measurement speed.
The software involves the control of high-frequency generator sweep, 3D geometry re-construction with time-of-flight signal and inner-frame phase-shift extraction with fast fourier transformation.
The result is submitted to IRMMW-THz international conference.
By taking advantage of THz wave transparency effect, skin movement by the conjunction of the heart is measured in real time with the radar while the participant has the clothes on.
This dramatically simplifies the labor required for heart rate measurement compared to conventional electrode-dependent measurement methods.
Until now, to inspect heart movement, the patient must first be half-naked, have multiple disposable electrodes attached, and lie in bed at rest. With this method, the heart movement can be measured with micro-order accuracy by simply sitting in front of the radar for a few tens of seconds while fully clothed.
End to end deep neural network classifier to distinguish 120 dog breed based on the photo. Over 20K+ images are imported from the Kaggle dataset and used for training. The model is transferred from pre-trained MobileNet V2 and customized for the problem. To visualize which portion of the image the computer is focusing on, each pixel's gradient based contribution score is calculated.
Face detection app with front-and-back end development including authentication management to provide unique save-data for each user.
Kanban board with drag-and-drop feature. You can add, edit, move and delete task on the board as you need.
Generate random quote with the help of API. Tweet button is implemented.
Web site which can toggle light / dark mode. The setting will be saved with localStorage.
Subtile animation is used as user scrolls the page to draw attention to the intended contents.
Email: 0512yuki@keio.jp