GuidedSAM‑Plume is an open‑source, expert‑guided framework designed for the interactive segmentation of volcanic plumes and clouds. (fact1)

The tool builds on SAM2 and uses its zero‑shot, spatial prompt‑based capabilities. Zero‑shot here means the framework works without task‑specific retraining. (fact2, fact5)

The authors validated GuidedSAM‑Plume on a challenging dataset using dedicated segmentation metrics and volcanology‑relevant descriptors to assess performance. (fact3)

With minimal user inputs, the framework produced a significant performance increase versus fully automatic dedicated solutions and preserved the reliability of derived volcanological parameters. (fact4, fact3)

Because GuidedSAM‑Plume does not require retraining, adoption barriers for volcanologists are lower. (fact5)

The project includes a re‑annotated dataset created under expert volcanologist guidance. That dataset is presented as a more standardized release. (fact6)

The framework is reported to be suitable for complex, diffuse plume‑shaped phenomena, which suggests potential applicability beyond volcanology to other environmental science uses. (fact7)

The code and accompanying dataset are available at the project repository. (fact8)