This repository includes codes of manuscript "Trans-SedNet: A Transformer-based dual-modal network for automatic segmentation of thin-section images of sandy sediments".
The Trans-SedNet is a ViT model that is designed for grain segmentation in geology, especially sedimentary geology.
- 'nex_cmx' and 'decoder' is the main code of Trans-SedNet, which are the encoder and decoder respectively.
- 'dual_segformer_w' is the backbone dual-modal SegFormer ViT model with window-attention.
- We also provide an example image dataset in 'dataset' folder in which one pair of XPL and PPL thin section images and their segmentation label are included.
The code is owned by Dongyu Zheng of Chengdu University of Technology, please feel free to contact the authors for further information.