Portfolio

Byungwoo Park

Vision AI Engineer · Autonomous Driving · 3D / 2D Computer Vision · Robotics

M.S. in Artificial Intelligence at Hanyang University.
Building perception systems where 3D data, deep learning, and robotics converge — from LiDAR-based object detection in autonomous driving research to 3D scan inspection pipelines deployed in industrial environments.

Email LinkedIn GitHub Google Scholar

Experience

2025.07 – 2025.12

Hypernoology

Vision AI Engineer

Senior Researcher / Team Lead

Developed 3D scan · alignment · inspection automation pipeline (UR10e + Zivid camera) from scratch — ROI ICP, RANSAC, anchor-based CAD alignment, multi-view stitching. Maintained ≤0.5 mm alignment error in production.
Built front-view object detection model for pedestrians, cyclists, and vehicles (YOLOv8, 8 MP camera); resolved class imbalance via oversampling + WeightedRandomSampler, achieving AP +4–6% on minority classes.
Industrial safety monitoring system — integrated 120k+ samples, auto-labeling pipeline, YOLO v11-pose + GRU-based anomaly detection, synthetic data generation for rare events.

2021.01 – 2022.04

Lotte Energy Materials

Process Engineer

Copper foil production & process management

Led copper recovery project from wastewater — ~18.5 tons/month recovered, ~₩240M/month cost savings.
Managed yield optimization and operator oversight for copper foil manufacturing line.

2019.10 – 2019.12

KITECH

Research Intern

Professional Service Robot Application Team

Developed tomato ripeness classification algorithm; built dataset collection and labeling pipeline.

Research & Publications

2024

ICRA

Fine-Grained Pillar Feature Encoding via Spatio-Temporal Virtual Grid for 3D Object Detection

IEEE International Conference on Robotics and Automation (ICRA) 2024
Y. Kim*, K. Park*, J. Koh, B. Park, J.W. Choi

Proposed a novel pillar encoding architecture that captures point cloud distributions across vertical, temporal, and horizontal axes. NDS +1.46% over baseline (latency: 63 → 69 ms).

ICTC

Enhancing Roadside 3D Object Detection Using Dynamic Pseudo HD Map

15th International Conference on ICT Convergence (ICTC) 2024
B. Park, Y. Kim, J. Kim, J.W. Choi

BEV voxel grid variance-based dynamic/static region classification to construct a lightweight Pseudo HD Map. mAP +3.3% over baseline (73.1 → 76.4).

KICS

LiDAR-based Multi-Object Detection in Unstructured Driving Environments

Korean Institute of Communications and Information Sciences (KICS) Summer 2024
B. Park, C. Kang, S. Ahn, J.W. Choi

2023 – 2024

Project

Driving Environment Data Stitching for Autonomous Vehicle Blind Spots

Korea Electronics Technology Institute (KETI) · Mar 2023 – Dec 2024

4-sensor LiDAR stitching pipeline at Pangyo Zero City intersection. Built detect–track–predict pipeline achieving FPS 31.4 (target: 20+) and mAP 60.1 (target: 60+).

Project

Robust Object Tracking in Off-road Environments for Convoy Driving

Hanwha Aerospace · Jan – Jun 2024

Sensor fusion pipeline (LiDAR + camera) for unmanned vehicle tracking in dense vegetation. Single-target tracking after 2D–3D match; MOTA up to 0.999, runtime ≤100 ms on embedded GPU (ADLINK EGX MXM-A4500).