Fewshot-VDT

Interactive Fusion and Correlation Network for Three-Modal Images Few-Shot Semantic Segmentation. As for the details, please refer to the paper:

• The paper has been accepted by IEEE SPL, you can refer to this link.

Usage

We have provided the prediction results of our network in this repository.

Code

The code for 1 shot model is in the folder ./1-shot and the code for 5 shot model is in the folder ./5shot. You can find the structure of our network in ./1-shot/model/mymodel.py and ./5-shot/model/mymodel.py.

usage

1. Download the dataset to ./VDT-2048-5i.
2. Prepare the environment.
3. Run the code:
   • training: 1-shot python./1-shot/train.py; 5-shot python./5-shot/train.py.
   • testing: 1-shot python./1-shot/test.py; 5-shot python./5-shot/test.py.

Dataset

The dataset can be downloaded from the link below:

• VDT-2048-5i