This PR improves the model card for `SAM-MT` by:
- Adding the `pipeline_tag: image-segmentation` metadata to help users discover the model.
- Linking the model card to the paper page ([SAM-MT: Real-Time Interactive Multi-Target Video Segmentation](https://huggingface.co/papers/2607.08688)).
- Adding links to the GitHub repository and project page.
- Including highlights, installation steps, inference commands, and citation info from the official repository.
