Improve model card for MedSAMix: Add pipeline tag, abstract, and detailed information

This PR significantly enhances the model card for `guinansu/MedSAMix-m-base` by:
- Adding the crucial `pipeline_tag: image-segmentation` to the metadata, which vastly improves the model's discoverability on the Hugging Face Hub.
- Enriching the content with the detailed abstract from the paper, providing a comprehensive overview of the model's methodology and results.
- Including the model architecture diagram from the GitHub repository for better visual understanding.
- Incorporating the checkpoint download links available on the project's GitHub page, making it easier for users to access the model weights in different formats.
- Updating the code link to also include the note about the code being cleaned and soon public, managing user expectations.

These changes ensure the model card is more informative and aligned with best practices for discoverability and user experience on the Hub.

README.md

@@ -1,13 +1,32 @@
 ---
+base_model:
+- facebook/sam-vit-base
 library_name: transformers
 tags:
 - medical
-base_model:
-- facebook/sam-vit-base
+pipeline_tag: image-segmentation
+license: mit
 ---
-Model MedSAMix-m （base）
 
-This model is based on the paper: https://arxiv.org/abs/2508.11032
+# MedSAMix: A Training-Free Model Merging Approach for Medical Image Segmentation
+
+This repository contains the `MedSAMix-m (base)` model, which is described in the paper [MedSAMix: A Training-Free Model Merging Approach for Medical Image Segmentation](https://arxiv.org/abs/2508.11032).
+
+Code: [https://github.com/podismine/MedSAMix](https://github.com/podismine/MedSAMix)
+Please note that the code is currently being cleaned and will be publicly released soon.
+
+## Abstract
+
+Universal medical image segmentation models have emerged as a promising paradigm due to their strong generalizability across diverse tasks, showing great potential for a wide range of clinical applications. This potential has been partly driven by the success of general-purpose vision models such as the Segment Anything Model (SAM), which has inspired the development of various fine-tuned variants for medical segmentation tasks. However, fine-tuned variants like MedSAM are trained on comparatively limited medical imaging data that often suffers from heterogeneity, scarce annotations, and distributional shifts. These challenges limit their ability to generalize across a wide range of medical segmentation tasks. In this regard, we propose MedSAMix, a training-free model merging method that integrates the strengths of both generalist models (e.g., SAM) and specialist models (e.g., MedSAM) for medical image segmentation. In contrast to traditional model merging approaches that rely on manual configuration and often result in suboptimal outcomes, we propose a zero-order optimization method to automatically discover optimal layer-wise merging solutions. Furthermore, for clinical applications, we develop two regimes to meet the demand of domain-specificity and generalizability in different scenarios by single-task optimization and multi-objective optimization respectively. Extensive evaluations on 25 medical segmentation tasks demonstrate that MedSAMix effectively mitigates model bias and consistently improves performance in both domain-specific accuracy and generalization, achieving improvements of 6.67% on specialized tasks and 4.37% on multi-task evaluations.
+
+<div align="center">
+  <img src="https://github.com/podismine/MedSAMix/raw/main/fig/model.png" alt="MedSAMix Model Architecture" width="60%">
+</div>
+
+## Checkpoint
+
+In addition, here we provide raw checkpoint and hugging face tensors:
 
+Pytorch raw checkpoint: [Here](https://drive.google.com/file/d/1RBsDZvFqJiAbbhnXTpSZs_uC-WKWrAJx/view?usp=sharing)
 
-Code: https://github.com/podismine/MedSAMix
+Hugging face: [Here](https://huggingface.co/guinansu/MedSAMix)