My research falls into three areas. Each section lists the related papers, project pages, and notes.

Plant architecture generated by a vision language model

Simulation and generative AI for plants

I use the Helios 3D plant simulator to generate training data for field images, and vision language models to build plant simulation models from images.

Working on a farm-ng ground robot

On-device AI and field robotics

I build and test robots, UAVs, and small devices that run AI models on board in the field. Examples include a GPS-guided spray robot for vineyards and a weed monitor that runs a quantized classifier on a Raspberry Pi AI camera.

VisTA-SR thermal super-resolution pipeline

Thermal and multimodal sensing

Low-cost thermal cameras make canopy temperature measurement affordable, but their images are low resolution and less accurate. I work on improving them with RGB images, and on combining thermal, RGB, and other sensors to estimate plant water status and traits.

Earlier work

At Seoul National University I worked on UAV remote sensing of upland crops (radiometric calibration, growth and biomass estimation) and built a weighing system for forest evapotranspiration, which was also used in a patented forest fire prevention system. From 2017 to 2021 I was a software engineer at A.I.MATICS, working on computer vision for driver monitoring, from embedded processors in vehicles to cloud processing on AWS.