Weijia (Allen) Wang | Geography & GeoAI

About

I am a thesis-based M.S. student in Geography at the University of Massachusetts Amherst, working with Dr. Hongyu Zhang on ethical and responsible GeoAI, with a focus on location privacy, geoprivacy, and place-based privacy.

My research connects GeoAI and Earth observation. I study the adaptation and evaluation of remote-sensing vision-language models, including their behavior under spatial distribution shift. I previously studied geospatial systems at Northern Illinois University, where I conducted research with Prof. Wei Luo.

Research interests

Responsible GeoAI & geoprivacy: Location privacy, place-based privacy, and the responsible use of geospatial data and AI.
Vision-language models for Earth observation: Parameter-efficient adaptation, model steering, and multi-label land-cover classification.
Spatial generalization & evaluation: Robustness across regions, sensors, seasons, and land-cover distributions.

Publications

Download BibTeX

2026 · Conference workshop paper

Steering Instruction-Tuned Vision-Language Models for Multi-Label Landcover Classification

Weijia Wang, Jichao Fang, Wei Luo

IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), 2026, pp. 1475–1483. 3rd Computer Vision for Earth Observation (CV4EO) Workshop.

Paper PDF

BibTeX

@InProceedings{Wang_2026_WACV,
  author = {Wang, Weijia and Fang, Jichao and Luo, Wei},
  title = {Steering Instruction-Tuned Vision-Language Models
           for Multi-Label Landcover Classification},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference
               on Applications of Computer Vision (WACV) Workshops},
  month = {March},
  year = {2026},
  pages = {1475-1483}
}

Presentations

2025 · Oral presentation

Parameter-Efficient Fine-Tuning of Vision-Language Models for High-Performance Earth Observation Tasks

Weijia Wang, Jichao Fang, Wei Luo

AGU Fall Meeting 2025, New Orleans, Louisiana.

Experience

Teaching Assistant

Sep. 2026–present

University of Massachusetts Amherst

Department of Earth, Geographic, and Climate Sciences.

Intern

Sep. 2026–present

Spatial Data Lab Internship Program · Remote

Research in GeoAI, Earth observation, spatial modeling, and geospatial analysis. Currently developing the research proposal under the guidance of Yongze Song.

Undergraduate Student Researcher

Aug. 2025–May 2026

Northern Illinois University

Research with Prof. Wei Luo on instruction-tuned vision-language models for Earth observation, including LoRA adaptation, model steering, and reproducible evaluation.

Academic service

Program Committee Member / Reviewer

Sep. 2026–present

4th Computer Vision for Earth Observation (CV4EO) Workshop at WACV 2027

Contributing to the workshop peer-review process at the intersection of computer vision, Earth observation, remote sensing, and GeoAI.

Education

University of Massachusetts Amherst

Aug. 2026–present

M.S. in Geography · Thesis-based

Department of Earth, Geographic, and Climate Sciences. Research with Dr. Hongyu Zhang.

Northern Illinois University

Aug. 2024–May 2026

B.S. in Earth, Atmosphere and Environment–Geospatial Systems

GPA: 3.897/4.000; summa cum laude. 2+2 dual degree program with Lanzhou Jiaotong University.

Lanzhou Jiaotong University

Sep. 2022–Jun. 2026

B.S. in Geographic Information Science

Training in geospatial science, GIS, and remote sensing.

Selected projects

Remote-sensing VLM adaptation

Instruction-tuned vision-language models for multi-label land-cover classification on BigEarthNet-S2, using parameter-efficient LoRA adaptation. Focus: model steering, reproducible evaluation, and failure analysis.

Ground-penetrating radar inversion

A TransUNet framework to reconstruct 2-D subsurface permittivity maps from GPR B-scan data, with physics-aware augmentation and composite loss design.

Remote-sensing language-model prototype

A domain-adapted VisualGLM-6B prototype for natural-language queries over Sentinel-2 imagery and Google Earth Engine assets, combining LoRA adaptation with an interactive Gradio interface.

Tools: Python, PyTorch, LoRA/QLoRA, Git, Google Earth Engine, ArcGIS Pro, QGIS, ENVI, ERDAS, and GeoDa.

Contact

For research inquiries and collaboration, please contact weijiawang@umass.edu.