News ·

FujiView appears in the WACV 2026 main proceedings

Bryce Bible presented the lab's multimodal scenic-visibility forecasting system at WACV in Tucson — predicting whether Mount Fuji will be visible with ~89% same-day accuracy.

Webcam frames of Mount Fuji with model visibility predictions overlaid.

FujiView: Multimodal Late-Fusion for Predicting Scenic Visibility — by PhD students Bryce Bible and Nehal Hasnaeen with Dr. Qi — was published in the main proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2026), held March 6–10 in Tucson, Arizona.

The paper introduces a forecasting framework built on more than 113,000 webcam images of Mount Fuji from 42 viewpoints, roughly 26,000 of them hand-labeled into five visibility classes, fused with concurrent and forecast weather data. Vision features dominate same-day prediction (~89% accuracy) while weather-forecast features carry the signal beyond one day (~84% next-day accuracy).

A public dataset, code release, and traveler-facing web app are in progress at fujiview.app. Read more on the project page.

All news