中文

SAM_TorchSharp is a project dedicated to exploring the feasibility of artificial intelligence development using .NET Core. Built upon TorchSharp and TorchSharp.PyBridge, this project focuses on porting the Segment-Anything (SAM) model from Python to the .NET Core ecosystem. SAM, an advanced image segmentation model, is supported in its sam_vit_b, sam_vit_l, and sam_vit_h variants, with successful integration of mobileSam. It introduces a versatile model loading mechanism that accepts weight files directly during initialization.

• Model Compatibility: Enables automatic loading of pretrained models by specifying weight file paths upon initialization.
• Exception Handling: Addresses issues encountered when attempting to load sam_vit_b_01ec64.pth, sam_vit_h_4b8939.pth, and sam_vit_l_0b3195.pth directly. As a workaround, it's advised to first save the model's state dictionary in Python using torch.save(model.state_dict(), "sam.pth"), which can then be seamlessly loaded within this project.
• .NET Core Integration: Leverages the cross-platform capabilities of .NET Core, expanding the horizons for AI application development.

• TorchSharp: A .NET binding to PyTorch, providing deep learning functionalities.
• TorchSharp.PyBridge: An extension for bridging Python and .NET environments, facilitating model migration.

Please refer to the INSTALL.md document for detailed installation instructions and environment setup guidelines. After cloning the project, ensure all dependencies are correctly installed and follow the provided instructions to set up your .NET Core environment.

We welcome all forms of contributions, including code submissions, bug reports, and documentation improvements. Consult the CONTRIBUTING.md file to learn how to get started.

This project is licensed under the MIT License, encouraging free use, modification, and distribution while preserving original authorship credits.

Special thanks go to:

• The TorchSharp team for providing powerful PyTorch bindings to .NET developers.
• The TorchSharp.PyBridge project for simplifying the complexities of cross-language model deployment.