Pipeline for Automated Snow Depth Extraction from Time-Lapse Camera Imagery

This repository contains code and sample data supporting the manuscript "Fully Automated Snow Depth Extraction from Time-Lapse Imagery Using Vision Foundation Models" submitted to Geophysical Research Letter for review. The pipeline combines SAM3 (Segment Anything Model 3) for zero-shot, text-prompted pole segmentation with Qwen3-VL-8B for automated image quality classification. No per-site annotation or model training is required. The method was evaluated across three NASA SnowEx campaigns (SNEX17, SNEX20, SNEX23). Input imagery is archived at the NASA National Snow and Ice Data Center (NSIDC). Corresponding author: Zhaocheng Wang ([e-mail]), School of Sustainable Engineering and the Built Environment, 777 East University Dr., WCPH Room 510, Arizona State University, Tempe, AZ, 85287-8704.