Add a paper or study
Copy any card above and replace the title, tag, summary, and link.
index / research
02 — Research
Research
Perception and planning work — published, in progress, and competition-tested. A few of the problems I've spent real time on.
Vision-Language · Published
+ New entry
Visual framing analysis with large VLMs
Evaluating multiple large vision-language models for analyzing visual framing in news images — toward scalable visual content analysis in mass communication. Lu, L., Wan, Z., Kwon, H., Kim, S. J., Kang, J., Abbas, L., Liu, J., & McLeod, D. M.
3D DetectionMono3DBox
Monocular 3D object detection that regresses oriented bounding boxes directly from 2D projections by adapting the YOLO architecture, with a convex-hull–based IoU. Outperforms baselines on car and pedestrian detection; nuScenes and BDD100k were reformatted to the KITTI schema to expand training. Pending publication.
Multi-Agent PlanningLeague of Robot Runners — 1st place
Placed 1st in the 2023 Amazon Robotics "League of Robot Runners" competition with a multi-agent coordination system for warehouse environments under strict compute limits — graph-based path optimization and priority queuing improved task completion by 194% on 150×500 grid maps.
DL Systemsmettle — tensor toolkit for Apple Silicon
A neural-network acceleration framework for Apple Silicon with custom Metal Shading Language kernels, reaching a 345% speedup over CPU baselines. Exploits unified memory to cut transfer overhead — with notable gains on LSTM/RNN — and ships a metal-cpp ndarray backend that drops into existing DL frameworks.