Case Study: Stanford University achieves rapid, high-precision satellite image labeling for urban air vehicle landing site detection with Labelbox

Stanford CS230 grad students research the next generation of AI-driven land urban air vehicles

Stanford University CS230 graduate students Andrew Denig and Seraj Desai needed roughly 5,000 satellite images segmented into landable vs. not‑landable regions for an urban air‑vehicle landing site project, but had only a short timeframe and could not afford the ~6 weeks it would take them to label everything themselves. They turned to Labelbox and its Workforce service to meet the rapid, high‑precision annotation needs for this challenging segmentation task.

Labelbox Workforce used the students’ 100 benchmark annotations and a one‑page guideline to begin labeling immediately, delivering more than 2,000 labeled images within a week and completing the ~5,000‑image dataset, exportable as JSON for easy model training. By offloading labeling to Labelbox, Andrew and Seraj gained seven full weeks to train and refine their networks, and the precisely traced contours produced by Labelbox measurably improved their model results beyond what the students could have achieved alone.

Stanford University

Andrew Denig

Master's Researchers