Image-Segmentation

Enhanced Convolutional Methods for Image Segmentation Study

This repository contains code and resources for studying enhanced convolutional methods in image segmentation architectures using PyTorch.

Objective

Explore image semantic segmentation architectures (Unet, AttenUnet, Unet++, Capsule-Net) by replacing standard convolutions with ten types of convolutions:

1. Standard
2. Spatially Separable
3. Gaussian Dynamic
4. Deformable
5. Adaptive Deformable
6. Standard Asymmetric
7. Asymmetric
8. Gaussian Dynamic Asymmetric
9. Deformable Asymmetric
10. Adaptive Deformable Asymmetric

Run the Code

1. Clone Repository: Clone this repository:

   git clone https://github.com/TechDaVinci/Image-Segmentation.git
   

2. Navigate to Project Directory: Go to project folder:

   cd Image-Segmentation
   

3. Run the Code:

   • Open main.ipynb in a Jupyter environment.
   • Run all cells.
   • Choose architecture and dataset.

4. Review Output:

   • Check output folder for results.
   • Explore segmentation outcomes.

Note

This repository enhances image segmentation with varied convolutions. Adapt code to your needs.

---