Processed UrbanSARFloods Dataset for Benchmarking PEFT Strategies with SAM

Résumé fourni par la source

The data were derived from the publicly available UrbanSARFloods dataset and were pre-processed for evaluating parameter-efficient fine-tuning (PEFT) strategies with the Segment Anything Model (SAM) for SAR-based flood mapping. The repository contains the processed dataset used for model development (datasetp) and four independent held-out test datasets (testweihui, testnovakakhovka, testjubba1, and testjubba2). These four test datasets correspond to the Weihui, Nova Kakhovka, and two independent Jubba image–label pairs that were excluded from model training and validation and used for independent evaluation in the study. The processed data are provided to facilitate reproducibility of the experiments and evaluation reported in the associated manuscript. Users of these data should also cite the original UrbanSARFloods dataset/publication, as the files provided here are derive

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.