Data and Code for "Mapping Tea Plantations in Hilly Terrain through Multiscale Context Modeling and Attention-Based Feature Refinement"

This dataset contains all data and code supporting the findings of the paper "Mapping Tea Plantations in Hilly Terrain through Multiscale Context Modeling and Attention-Based Feature Refinement". The archive includes:1. Complete code implementations of four network variants: Swin-UNet baseline, Swin-UNet+CBAM, Swin-UNet+ASPP, and Swin-UNet+CBAM+ASPP. The code includes data preprocessing, model training, evaluation scripts, and a requirements.txt file for dependency installation. 2. Preprocessed 224×224 image patches with corresponding binary tea plantation masks, split into training (70%), validation (10%), and test (20%) sets. The data are derived from 0.5 m Jilin-1 multispectral imagery acquired across 15 sites in Lin'an District, Hangzhou, China, during spring, summer, and autumn 2023. All materials are released under a CC BY 4.0 license. The code was developed using PyTorch and Swin-UNet architecture. For any questions regarding this dataset, please contact the corresponding author: [e-mail].