Haruya.
I'm a PhD candidate with a focus on Computer Vision and Deep Learning.
A passionate research engineer motivated to invent and have fun while doing it.
Hi, nice to meet you. I'm
About Me
My research interests are broadly in computer vision as well as a bit of robotics and reinforcement learning. I worked on projects with Visual SLAM, multi-object tracking, action recognition, and semantic segmentation.
I love learning new technologies, drinking coffee, and gaming (guilty!).
Contact me! Let’s go for a coffee!Here are a few technologies I've been working with recently:
- PyTorch
- neovim
- Hugging Face
- OpenGL
- Unity
- Rust
Experience
Research in active vision for embodied agents.
- Research in Visual SLAM, Deep Learning, and Reinforcement Learning under the supervision of PhD. Ken Sakurada.
- Trained and benchmarked active vision agents for robust visual SLAM.
- Assisted in more than 8 workshops in Tokyo relating to deep learning and CUDA GTC Japan DLI.
- Translated labs to be used for tutorial of Digits framework.
- Became a certified CUDA developer.
- Developed an iOS appliacation called “Graffity” which lets users create and share AR experiences. The app made top 15 on the Apple App Store and was reviewed on famous news outlets:
- One of my main responsibilities were to design and impliment AR UI/UX for the camera using ARKit and Vision frameworks.
- Executed rapid hypothesis testing for market research with heavy emphasis on 0 to 1 development.
- Researched and implemented techniques for fast and robust re-localization for seamless user experience.
- Developed object tracking application using Intel’s Computer Vision SDK.
- Developed and documented clasic object and face detection algorithms using C++ to be included in the SDK.
- Published documentation for Japanese clients and users for the SDK.
- Was part of the Brain Portal (Dept. of Media Technology Laboratory) team which specializes in supporting hardware startups around the world.
- My responsibilities were to develop software to gather and analyze possible clients and manufacturers by automating the search process using web crawlers.
- Utilized Python for data mining and analysis, and developed pipelines for Google Sheets API.
- My main responsibilities were to design client application aimed at fashion retail shops to visualize sales and to develop a protoype using Vue.js framework.
Education
2020 - Present
PhD in Computer Science
Keio University
- Supervised by Prof. Yoshimitsu Aoki (Aoki Media Lab)
2018 - 2020
MA in Computer Science (Integrated Design Engineering)
Keio University
- Supervised by Prof. Yoshimitsu Aoki (Aoki Media Lab)
- Thesis: “Active Neural SLAM: Deep Reinforcement Learning for Visual SLAM”
2014 - 2018
BA in Electrical Engineering and Informatics
Keio University
GPA:
3.72 out of 4.0
- Supervised by Prof. Yukitoshi Sanada (Sanada Lab)
- Thesis: “Effect of Joint Detection on System Throughput in Distributed Antenna Network”
2011 - 2014
High School
International Christian University High SchoolProjects
PyTorch
Numpy
Python
Equilib
Equirectangular (360/panoramic) image processing library for Python (PyTorch and numpy) with minimal dependencies.
Python
C/C++
PyEdgeEval
Edge detection benchmark and toolkit.
Python
PyTorch
Multi-Object Tracking in Crowded Scenes
In this project, I combined modern state-of-the-art re-identification models and modeling techniques on the basic tracking-by-detection framework and benchmark them on heavily occluded scenes to understand their effect.
Python
C/C++
three.js
Open3D
Human-Object Map
Visualize and understand human behavior by using human-object maps (HOMs). Accompanies my MVA 2019 paper.
MATLAB
5G Distributed Antenna System Simulation
I performed throughput analysis for distuted antenna system (DAS) which is proposed in the 5G communication protocol. In DAS, mobile terminals are closer to the desired antennas, increasing the frequency utilization efficiency, but also increases inter-channel inference. I proposed joint likelihood detection to increase throughput in a multi-user scheduling scenario.
Swift
ARKit
Vision
Graffity
AR social media application that lets user share AR experience with people around the world. iOS application created with Swift and ReactNative.
Python
AWS
Scholar.ai
AI papers categorized, filtered, and updated everyday. Developed while working for AINOW (DIP).
Achievements
Research Fellowship for Young Scientists DC1
3 Year Research Grant (3.5 million JPY) + 200,000 JPY per month.
JEES Mitsui Corporation Science and Technology Scholarship
1 Year Scholarship 100,000 JPY per month.
JGC-S Scholarship
300,000 JPY
Keio University Scholarship
500,000 JPY
JASSO Scholarship (Top 30% Ecellent Graduate School Students)
800,000 JPY
Young Researchers ́ Encouragement Award
IEEE VTS Japan Chapter 2018
Get in Touch
My inbox is always open. Whether you have a question or just want to say hi, I’ll try my best to get back to you!