Katsuya Hyodo
Research Engineer at CyberAgent AI Lab
Nagoya, Japan
Summary
🤩
Rockstar🎓
Top SchoolKatsuya Hyodo is a Research Engineer based in Nagoya with eight years of software engineering experience, currently contributing to CyberAgent AI Lab after a six-year tenure at TOYOTA INDUSTRIES IT SOLUTIONS. He specializes in ML model optimization, conversion, and deployment—especially across TensorFlow, PyTorch, ONNX, OpenVINO, and edge runtimes—maintaining a popular model-zoo focused on cross-framework interoperability and quantization. An Intel Software Innovator program member and active hobbyist programmer, he has practical experience bringing models like MobileNet and ResNet to resource-constrained devices including EdgeTPU and Raspberry Pi. With a background in political economy, he combines a broad analytical perspective with hands-on engineering to translate research models into production-ready, efficient deployments.
8 years of coding experience
6 years of employment as a software developer
Political Economy, Political Economy at Shiga University
Stackoverflow
36reputation
6kreached
4answers
0questions
Github Skills (20)
python10
machine-learning10
openvino10
deep-learning10
tensorflow10
model-optimization10
object-detection10
computer-vision10
tflite10
model-conversion10
yolov410
opencv8
cpu8
raspberry-pi6
depth6
Programming languages (11)
Github contributions (5)
PINTO0309/PINTO_model_zoo
Dec 2019 - Jan 2023
A repository for storing models that have been inter-converted between various frameworks. Supported frameworks are TensorFlow, PyTorch, ONNX, OpenVINO, TFJS, TFTRT, TensorFlowLite (Float32/16/INT8), EdgeTPU, CoreML.
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 2216 commits, 126 PRs in 3 years 2 months
Contributions summary:Katsuya's contributions focused on storing machine learning models that have been inter-converted between various frameworks. They added and modified scripts for processing different deep learning model versions, including MobileNet and ResNet, by converting them to SavedModel format. They also added various scripts for quantizing the models to various precisions (float16, int8) and added support for EdgeTPU. The user seems to be involved in model optimization, conversion, and deployment across different frameworks.
tensorflowjstf-trtmodelfloat32inter
PINTO0309/OpenVINO-YoloV3
Dec 2018 - Dec 2021
YoloV3/tiny-YoloV3+RaspberryPi3/Ubuntu LaptopPC+NCS/NCS2+USB Camera+Python+OpenVINO
Role in this project:ML Engineer
ML EngineerContributions:1 release, 198 commits, 3 PRs in 3 years
Contributions summary:Katsuya's commits primarily involve modifications to `utils.py` and `convert_weights_pb.py`, suggesting involvement in model conversion and utility functions. These changes relate to loading weights, freezing graphs, and preparing models, all crucial steps in deploying a machine-learning model within the OpenVINO framework, as the repository description indicates. Further commits show work on downloading pre-trained models and configuring the output nodes, showing strong involvement with model deployment and the OpenVINO toolset.
cpuraspberry-pipythondeep-learningcamera
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.