Add comprehensive dataset card for SonoGym_lerobot_dataset

README.md

@@ -0,0 +1,67 @@
+---
+task_categories:
+- robotics
+license: cc-by-nc-4.0
+library_name: lerobot
+tags:
+- simulation
+- surgical-robotics
+- ultrasound
+- reinforcement-learning
+- imitation-learning
+- medical-imaging
+- isaac-lab
+---
+
+# SonoGym: Expert Dataset for Robotic Ultrasound Tasks
+
+This repository contains the expert datasets collected for training imitation learning policies as part of the [SonoGym: High Performance Simulation for Challenging Surgical Tasks with Robotic Ultrasound](https://huggingface.co/papers/2507.01152) project.
+
+**Project Page:** [https://sonogym.github.io/](https://sonogym.github.io/)
+**SonoGym GitHub Repository:** [https://github.com/SonoGym/SonoGym](https://github.com/SonoGym/SonoGym)
+
+## Overview of SonoGym
+
+SonoGym is a scalable simulation platform for complex robotic ultrasound tasks built on NVIDIA IsaacLab. It enables parallel simulation across tens to hundreds of environments and supports realistic, real-time simulation of US data from CT-derived 3D models of anatomy. The framework facilitates the training of deep reinforcement learning (DRL) and imitation learning (IL) agents for relevant tasks in robotic orthopedic surgery, integrating common robotic platforms and orthopedic end effectors.
+
+## Dataset Details
+
+This `SonoGym_lerobot_dataset` specifically provides expert trajectories collected within the SonoGym simulation environment. These trajectories are designed for use with the [lerobot library](https://github.com/huggingface/lerobot) to train imitation learning policies for surgical and navigation tasks. This dataset is not necessary for training pure reinforcement learning agents.
+
+The dataset includes expert demonstrations collected under various simulation settings:
+
+*   `Isaac-robot-US-guidance-v0-single`: Ultrasound guidance, single patient, model-based US simulation.
+*   `Isaac-robot-US-guidance-v0-single-net`: Ultrasound guidance, single patient, learning-based US simulation.
+*   `Isaac-robot-US-guidance-5-models-v0`: Ultrasound guidance, single patient, using 4 learning-based US simulation networks.
+*   `Isaac-robot-US-guided-surgery-v0-single-new`: Ultrasound-guided surgery, single patient, model-based US simulation.
+*   `Isaac-robot-US-guided-surgery-v0-single-net-new`: Ultrasound-guided surgery, single patient, learning-based US simulation.
+*   `Isaac-robot-US-guided-surgery-v0-5-net`: Ultrasound-guided surgery, single patient, using 4 learning-based US simulation networks.
+*   `Isaac-robot-US-guided-surgery-v0-5`: Ultrasound-guided surgery, 5 patients, model-based US simulation.
+
+## Sample Usage
+
+To utilize this dataset for training imitation learning policies (e.g., ACT or Diffusion Policy) with the `lerobot` library, follow these steps:
+
+1.  **Download the dataset:**
+    It is recommended to download the dataset and place it in `SonoGym/lerobot-dataset` within your SonoGym project directory for seamless integration.
+
+2.  **Train an Imitation Learning Agent:**
+    Ensure you have `lerobot` installed. You can then use the training scripts provided within the SonoGym codebase:
+    ```bash
+    python /path-to-lerobot/lerobot/scripts/train.py --config_path=workflows/lerobot/train_surgery_{method}_cfg.json
+    ```
+    Replace `{method}` with either `'diffusion'` or `'act'`.
+    **Important:** Remember to update the `"dataset": "root"` entry in the relevant `train_surgery_{method}_cfg.json` configuration file to the absolute path of your downloaded dataset, for example: `"/path-to-repo/SonoGym/lerobot-dataset/Isaac-robot-US-guidance-v0-single-net"`.
+
+## Citation
+
+If you find our dataset useful for your research and applications, please cite the associated paper:
+
+```bibtex
+@inproceedings{cao2024sonogym,
+      title={SonoGym: High Performance Simulation for Challenging Surgical Tasks with Robotic Ultrasound},
+      author={Yun-Hsuan Cao and Mengchen Zhang and Gordon Wetzstein and Ryan Po and Ru-Yuan Zhang and Peter X. K. Lu},
+      year={2024},
+      url={https://huggingface.co/papers/2507.01152},
+}
+```