Data and Code for River Network Extraction in the Three Gorges Reservoir Area Using a Transformer-Enhanced U-Net and High-Resolution Jilin-1 Satellite Imagery

This repository contains the dataset, source code, trained model weights, and experimental results supporting the study: “Accurate Extraction of River Networks in the Three Gorges Reservoir Area Using a Transformer-Enhanced U-Net and High-Resolution Jilin-1 Satellite Data”. The study focuses on river network extraction in the Three Gorges Reservoir Area, Chongqing, China, using a Transformer-enhanced U-Net (TransUNet) and high-resolution Jilin-1 satellite imagery. The repository is designed to support reproducibility and includes all necessary components for model training, testing, and evaluation. Contents include:Source code (model implementation, training and inference scripts)Sample datasets (DRIVE and equivalent structured data)Trained model weights (vit_seg.pt)Pretrained backbone weights (ImageNet-based ViT model)Prediction results (binary river extraction maps and outputs)Evaluation metrics (accuracy, precision, recall, F1-score, IoU, loss curves) Study Area:Three Gorges Reservoir Area, Chongqing, China. Data Availability:Due to licensing restrictions, the original high-resolution Jilin-1 satellite imagery cannot be publicly shared. Only derived data products, sample datasets, trained models, and code are provided in this repository to ensure reproducibility. Software Environment:Python 3.8.1 PyTorch 1.12.1 CUDA Toolkit 11.6.2 Dependencies are listed in requirements.txt. Usage:Run the main program:python run.py Optional GUI mode:python GUI.py License:This repository is distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.