Update README.md

README.md

@@ -1,3 +1,85 @@
 ---
 license: apache-2.0
 ---
+# Person2Drive: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving
+
+This repository hosts the dataset and benchmark resources for the ECCV 2026 paper:
+
+**Driving like yourself: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving**
+
+## Overview
+
+Person2Drive is a closed-loop personalized end-to-end autonomous driving benchmark collected in the CARLA simulator. The dataset is designed to support research on human driving style modeling, driver-level personalization, and closed-loop evaluation for end-to-end autonomous driving.
+
+Unlike conventional driving datasets that mainly focus on generic driving behavior, Person2Drive provides driver-level driving records, enabling the study of how different human drivers behave under the same or similar driving scenarios.
+
+## Release Status
+
+This Hugging Face repository is the stable public access page for the dataset and benchmark.
+
+The dataset files are currently being uploaded and organized. The repository will be progressively updated with the remaining data files, metadata, benchmark documentation, and usage instructions before the ECCV 2026 camera-ready deadline.
+
+Files already available in this repository are part of the public release.
+
+## Dataset Content
+
+The dataset contains human driving records collected in closed-loop CARLA environments. The full release is organized around multiple human drivers and multiple driving routes.
+
+Each driving record may include:
+
+* ego-vehicle states;
+* control signals, such as steering, throttle, and brake;
+* ego trajectories;
+* route and navigation information;
+* selected sensor data used for end-to-end driving models;
+* metadata for driver-level and route-level organization.
+
+The dataset supports both open-loop behavior analysis and closed-loop personalized driving evaluation.
+
+## Dataset Structure
+
+The dataset is organized by driver, route, and episode. A typical structure is:
+
+```text
+Person2Drive/
+├── driver_xxx/
+│   ├── route_xxx/
+│   │   ├── episode_xxx/
+│   │   │   ├── measurements/
+│   │   │   ├── sensor_data/
+│   │   │   ├── trajectories/
+│   │   │   └── metadata/
+│   │   └── ...
+│   └── ...
+├── metadata/
+├── benchmark/
+└── README.md
+```
+
+The exact file structure may vary slightly across released subsets and will be documented in detail as the upload is completed.
+
+## Benchmark Usage
+
+Person2Drive is intended for evaluating personalized end-to-end autonomous driving methods. The benchmark focuses on whether a model can adapt to individual driving styles while maintaining safe closed-loop driving performance.
+
+Typical evaluation settings include:
+
+* driver-level personalization;
+* route-level generalization;
+* trajectory prediction quality;
+* driving style similarity;
+* closed-loop driving performance.
+
+Detailed benchmark scripts, evaluation protocols, and usage instructions will be added to this repository.
+
+## Notes on Data Organization
+
+Driver identities are represented using anonymized driver identifiers. The dataset is intended for research use in personalized autonomous driving, driving behavior modeling, and closed-loop simulation-based evaluation.
+
+## Citation
+
+The citation information will be added after the ECCV 2026 camera-ready version is finalized.
+
+## Contact
+
+For questions about the dataset, please contact the authors of the paper.