Enhance GoalFlow model card with comprehensive details and metadata

README.md

@@ -1,6 +1,152 @@
 ---
 license: apache-2.0
+pipeline_tag: robotics
+library_name: diffusers
+tags:
+  - autonomous-driving
+  - trajectory-generation
+  - flow-matching
 ---
-The repository contains the model described in the paper [GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving](https://arxiv.org/abs/2503.05689)
 
-Code: https://github.com/YvanYin/GoalFlow
+<p align="center">
+    <img alt="goalflow_logo" src="https://github.com/YvanYin/GoalFlow/raw/main/assets/goalflow_logo.png" width="500">
+    <h3 align="center"><a href="https://arxiv.org/abs/2503.05689">Paper</a> | <a href="https://drive.google.com/drive/folders/1iWsPwpqM4WaUVVRZU3xIMPdOaJVB2Kub?usp=drive_link">Weight</a> | <a href="https://zebinx.github.io/HomePage-of-GoalFlow/">ProjectPage</a> | <a href="https://github.com/YvanYin/GoalFlow">Code</a> </h3>
+</p>
+
+<br/>
+
+> [**GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation
+in End-to-End Autonomous Driving**](https://arxiv.org/abs/2503.05689)  <br>
+> [Zebin Xing](https://github.com/ZebinX)<sup>1,2*</sup>, [Xingyu Zhang]()<sup>2*</sup>, [Yang Hu]()<sup>1,2</sup>, [Bo Jiang]()<sup>4,2</sup>, [Tong He](https://tonghe90.github.io/)<sup>5</sup>, [Qian Zhang]()<sup>2</sup>, [Xiaoxiao Long](https://www.xxlong.site/)<sup>3</sup>, [Wei Yin](https://yvanyin.net/)<sup>2✝</sup>  <br>
+> <sup>1</sup> University of Chinese Academy of Sciences, <sup>2</sup> Horizon Robotics, <sup>3</sup> Nanjing University, <sup>4</sup> Huazhong University of Science & Technology, <sup>3</sup> Shanghai AI Laboratory  <br>
+> <br>
+> Computer Vision and Pattern Recognition (CVPR), 2025 <br>
+>
+This is the official repo of 'GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving (CVPR 2025)'. GoalFlow achieved PDMS of 90.3, significantly surpassing other baselines. Compared with other diffusion-policy-based methods, our approach requires only a single denoising step to obtain excellent performance.
+
+## Abstract
+We propose GoalFlow, an end-to-end autonomous driving method for generating high-quality multimodal trajectories. In autonomous driving scenarios, there is rarely a single suitable trajectory. Recent methods have increasingly focused on modeling multimodal trajectory distributions. However, they suffer from trajectory selection complexity and reduced trajectory quality due to high trajectory divergence and inconsistencies between guidance and scene information. To address these issues, we introduce GoalFlow, a novel method that effectively constrains the generative process to produce high-quality, multimodal trajectories. To resolve the trajectory divergence problem inherent in diffusion-based methods, GoalFlow constrains the generated trajectories by introducing a goal point. GoalFlow establishes a novel scoring mechanism that selects the most appropriate goal point from the candidate points based on scene information. Furthermore, GoalFlow employs an efficient generative method, Flow Matching, to generate multimodal trajectories, and incorporates a refined scoring mechanism to select the optimal trajectory from the candidates. Our experimental results, validated on the Navsim, demonstrate that GoalFlow achieves state-of-the-art performance, delivering robust multimodal trajectories for autonomous driving. GoalFlow achieved PDMS of 90.3, significantly surpassing other methods. Compared with other diffusion-policy-based methods, our approach requires only a single denoising step to obtain excellent performance.
+
+<br/>
+
+## News
+*   **`20 Mar, 2025`:** We released our paper on [arXiv](https://arxiv.org/abs/2503.05689). Code is coming soon.
+*   **`27 Feb, 2025`:** GoalFlow was accepted at [CVPR](https://cvpr.thecvf.com/Conferences/2025) !
+
+<br/>
+
+## Introduction
+> In autonomous driving, multiple optimal trajectories exist, like overtaking or following. (1) Traditional methods efficiently generate safe trajectories but struggle with multimodal ones. (2) Generative methods like diffusion models capture multimodal distributions but require heavy hardware and prior information. We propose GoalFlow, a goal-point-based method that guides trajectory planning. With a map-free evaluation and an efficient diffusion variant, Flow Matching, we reduce inference steps, achieving superior performance with just one denoising step.
+
+<div align="center">
+<img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/main_fig.png" />
+</div>
+
+## Visualization
+
+### Comparison with Other Methods
+❌ indicates that the trajectory results in a collision or goes beyond the drivable area, while ✅ represents a safe trajectory. The orange points are optimal goal points evaluated by the Goal Constructor, while the blue and yellow points correspond to samples from the vocabulary.
+<div align="center">
+    <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visualization.png" />
+</div>
+
+### Driving Vedios
+Driving Vedios generated by GoalFlow.
+<table style="width: 100%; table-layout: fixed;">
+    <tr>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/cf12097663665430.gif" style="width: 100%; height: auto;">
+        </td>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/25b719c231d85e56.gif" style="width: 100%; height: auto;">
+        </td>
+    </tr>
+    <tr>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/1a1fbb255ec55813.gif" style="width: 100%; height: auto;">
+        </td>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/676880c7e31252c0.gif" style="width: 100%; height: auto;">
+        </td>
+    </tr>
+    <tr>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/d2440edd19d954b5.gif" style="width: 100%; height: auto;">
+        </td>
+        <td style="width: 50%; text-align: center;">
+            <img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/fb0a26a28ec359ce.gif" style="width: 100%; height: auto;">
+        </td>
+    </tr>
+</table>
+
+### Goal Point Distribution
+From top to down, they are respectively the distributions of DAC, distance, and the final score. The points with warmer color have higher score.
+<table style="width: 100%; table-layout: fixed; text-align: center;">
+    <tr>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/dac_scores/0a44947ca9e85579.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/dac_scores/2a06b778a64b545e.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/dac_scores/7abf60c1594953cf.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/dac_scores/1db7c81f96855ce9.png" style="width: 100%; height: auto;"></td>
+    </tr>
+    <tr>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/im_scores/0a44947ca9e85579.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/im_scores/2a06b778a64b545e.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/im_scores/7abf60c1594953cf.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/im_scores/1db7c81f96855ce9.png" style="width: 100%; height: auto;"></td>
+    </tr>
+    <tr>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/final_scores/0a44947ca9e85579.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/final_scores/2a06b778a64b545e.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/final_scores/7abf60c1594953cf.png" style="width: 100%; height: auto;"></td>
+        <td><img src="https://github.com/YvanYin/GoalFlow/raw/main/assets/visual_goal_point/final_scores/1db7c81f96855ce9.png" style="width: 100%; height: auto;"></td>
+    </tr>
+</table>
+
+
+## Results
+Planning results on the proposed **NAVSIM** **Test** benchmark. Please refer to the [paper](https://arxiv.org/abs/2503.05689) for more details.
+
+| Method           | S<sub>NC</sub> ↑ | S<sub>DAC</sub> ↑ | S<sub>TTC</sub> ↑ | S<sub>CF</sub> ↑ | S<sub>EP</sub> ↑ | S<sub>PDM</sub> ↑ |
+|-----------------|-----------------|-----------------|-----------------|-----------------|-----------------|-----------------|
+| Constant Velocity | 68.0 | 57.8 | 50.0 | 100 | 19.4 | 20.6 |
+| Ego Status MLP   | 93.0 | 77.3 | 83.6 | 100 | 62.8 | 65.6 |
+| LTF             | 97.4 | 92.8 | 92.4 | 100 | 79.0 | 83.8 |
+| TransFuser      | 97.7 | 92.8 | 92.8 | 100 | 79.2 | 84.0 |
+| UniAD          | 97.8 | 91.9 | 92.9 | 100 | 78.8 | 83.4 |
+| PARA-Drive     | 97.9 | 92.4 | 93.0 | 99.8 | 79.3 | 84.0 |
+| **GoalFlow (Ours)** | **98.4** | **98.3** | **94.6** | **100** | **85.0** | **90.3** |
+| *Human<sup>‡</sup>* | *100* | *100* | *100* | *99.9* | *87.5* | *94.8* |
+
+
+## Getting started
+- [Download Datasets of NAVSIM](https://github.com/autonomousvision/navsim/blob/main/docs/install.md)
+- [Preparation of GoalFlow Environment](https://github.com/YvanYin/GoalFlow/blob/main/docs/install.md)
+- [Evaluation](https://github.com/YvanYin/GoalFlow/blob/main/docs/test.md)
+- [Training](https://github.com/YvanYin/GoalFlow/blob/main/docs/train.md)
+
+## Contact
+If you have any questions or suggestions, please feel free to open an issue or contact us (xzebin@bupt.edu.cn).
+
+## Acknowledgement
+<p>1. We have gained valuable insights from <a href="https://arxiv.org/abs/2406.06978" target="_blank">Hydra-MDP</a>, which provided many inspiring ideas referenced in our work.</p>
+<p>2. We referred to an excellent GitHub project, <a href="https://github.com/autonomousvision/tuplan_garage" target="_blank">tuplan garage</a>, and incorporated aspects of its page design.</p>
+
+<p>3. GoalFlow is also greatly inspired by the following outstanding contributions to the open-source community:</p>
+<ul>
+    <a href="https://github.com/autonomousvision/navsim" target="_blank">NAVSIM</a> | <a href="https://github.com/autonomousvision/transfuser" target="_blank">TransFuser</a> | <a href="https://github.com/hustvl/VAD" target="_blank">Diffusion-ES</a> | <a href="" target="_blank">VAD-v2</a>
+</ul>
+
+
+
+## Citation
+If you find GoalFlow useful, please consider giving us a star &#127775; and citing our paper with the following BibTeX entry.
+
+```BibTeX
+@article{xing2025goalflow,
+	  title={GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving},
+	  author={Xing, Zebin and Zhang, Xingyu and Hu, Yang and Jiang, Bo and He, Tong and Zhang, Qian and Long, Xiaoxiao and Yin, Wei},
+	  journal={arXiv preprint arXiv:2503.05689},
+	  year={2025}}
+```
+
+<p align="right">(<a href="#top">back to top</a>)</p>