8 papers
Beyond Simulation: Retain-and-Repair Neural Operators for Real-World Adaptation
Woojin Cho, Junghwan Park
Neural operators increasingly benefit from pretraining on numerical simulations, yet adapting them for real-world prediction remains challenging. We introduce the Retain-and-Repair…
Summarize First, Download Later: Onboard VLMs for Bandwidth-Efficient Earth Observation
Junghwan Park, Sangcheol Sim, Woojin Cho +1
Modern Earth observation (EO) satellites carry increasingly advanced sensors that produce vast volumes of high-resolution, multispectral data, yet downlink capacity remains a criti…
ELMZip: Onboard Satellite Image Compression via Extreme Learning Machines for Efficient Downlink
Woojin Cho, Junghwan Park, Sangcheol Sim +3
The acquisition of multispectral imagery via small satellites (e.g., CubeSats) presents significant data downlink challenges due to high data volumes and restricted communication w…
Few-Shot Open-Vocabulary Remote Sensing Segmentation via Textual Inversion
Junhyuk Heo, Junghwan Park
Open-vocabulary segmentation labels arbitrary categories from a text query without per-class training, yet on remote sensing imagery it underperforms on categories it handles relia…
FROST: Training-Free Few-Shot Segmentation with Frozen Features and Nonparametric Statistics
Junghwan Park
Few-shot segmentation asks a model to delineate a target class in a query image from only a handful of annotated examples, a setting most acute in remote sensing, where labels are…
Escaping Spectral Bias without Backpropagation: Fast Implicit Neural Representations with Extreme Learning Machines
Woojin Cho, Junghwan Park
Training implicit neural representations (INRs) to capture fine-scale details typically relies on iterative backpropagation and is often hindered by spectral bias when the target e…
Basis-Oriented Low-rank Transfer for Few-Shot and Test-Time Adaptation
Junghwan Park, Woojin Cho, Junhyuk Heo +2
Adapting large pre-trained models to unseen tasks under tight data and compute budgets remains challenging. Meta-learning approaches explicitly learn good initializations, but they…
PIANO: Physics-informed Dual Neural Operator for Precipitation Nowcasting
Seokhyun Chin, Junghwan Park, Woojin Cho
Precipitation nowcasting, key for early warning of disasters, currently relies on computationally expensive and restrictive methods that limit access to many countries. To overcome…