Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: pytorch
|
| 4 |
+
tags:
|
| 5 |
+
- 3d-object-detection
|
| 6 |
+
- cooperative-perception
|
| 7 |
+
- autonomous-driving
|
| 8 |
+
- lidar
|
| 9 |
+
- camera
|
| 10 |
+
- v2x
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# CooperScene: Multi-Modal Cooperative Autonomy Benchmark with C-V2X Communication Characterization
|
| 14 |
+
|
| 15 |
+
[](https://cisl.ucr.edu/CooperScene)
|
| 16 |
+
[](https://github.com/UCR-CISL/CooperScene)
|
| 17 |
+
[](https://huggingface.co/cisl-hf/CooperScene)
|
| 18 |
+
[](https://pytorch.org/get-started/locally/)
|
| 19 |
+
[](https://github.com/UCR-CISL/CooperScene/blob/main/LICENSE)
|
| 20 |
+
|
| 21 |
+
## Introduction
|
| 22 |
+
|
| 23 |
+
🚗 This repository hosts the **model configs and pre-trained checkpoints** for
|
| 24 |
+
[CooperScene](https://cisl.ucr.edu/CooperScene) — the first real-world,
|
| 25 |
+
multi-agent, multi-modal cooperative autonomy dataset with C-V2X communication
|
| 26 |
+
characterization (three connected vehicles + one roadside unit, across
|
| 27 |
+
intersections, highway ramps, and parking areas).
|
| 28 |
+
|
| 29 |
+
🚀 All training and inference code is open-sourced. See the
|
| 30 |
+
[project page](https://cisl.ucr.edu/CooperScene) and the
|
| 31 |
+
[GitHub repo](https://github.com/UCR-CISL/CooperScene) for details.
|
| 32 |
+
|
| 33 |
+
💬 We welcome feedback and look forward to your comments!
|
| 34 |
+
|
| 35 |
+
## What's here
|
| 36 |
+
|
| 37 |
+
Each model has its config and matching checkpoint together under
|
| 38 |
+
`configs/<model>/`:
|
| 39 |
+
|
| 40 |
+
| Cooperative detectors | BEVFusion |
|
| 41 |
+
|---|---|
|
| 42 |
+
| `cobevt` | `bevfusion_single_lidar` |
|
| 43 |
+
| `cosdh` | `bevfusion_single_lidarcam` |
|
| 44 |
+
| `ermvp` | `bevfusion_coop_lidar` |
|
| 45 |
+
| `v2vam` | `bevfusion_coop_lidarcam` |
|
| 46 |
+
| `v2vnet` | |
|
| 47 |
+
| `v2xvit` | |
|
| 48 |
+
|
| 49 |
+
All models run on a unified mmengine pipeline (`proj_first=True`, same global-sort
|
| 50 |
+
BEV/3D polygon-IoU AP @ 0.3 / 0.5 / 0.7), so the numbers are directly comparable.
|
| 51 |
+
|
| 52 |
+
## Download
|
| 53 |
+
|
| 54 |
+
```bash
|
| 55 |
+
pip install -U huggingface_hub
|
| 56 |
+
hf download cisl-hf/CooperScene --local-dir assets
|
| 57 |
+
# -> assets/configs/<model>/{<model>.py, <model>.pth}
|
| 58 |
+
```
|
| 59 |
+
|
| 60 |
+
## Usage
|
| 61 |
+
|
| 62 |
+
Clone the [code repo](https://github.com/UCR-CISL/CooperScene), then evaluate or
|
| 63 |
+
train with a downloaded config + checkpoint:
|
| 64 |
+
|
| 65 |
+
```bash
|
| 66 |
+
# evaluate (test split by default)
|
| 67 |
+
python tools/test.py assets/configs/ermvp/ermvp.py assets/configs/ermvp/ermvp.pth
|
| 68 |
+
|
| 69 |
+
# train (warm-start from a checkpoint, optional)
|
| 70 |
+
python tools/train.py assets/configs/ermvp/ermvp.py
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
See the [GitHub README](https://github.com/UCR-CISL/CooperScene) for data
|
| 74 |
+
preparation and the Docker workflow.
|
| 75 |
+
|
| 76 |
+
## Related links
|
| 77 |
+
|
| 78 |
+
🌐 Website: [https://cisl.ucr.edu/CooperScene](https://cisl.ucr.edu/CooperScene)
|
| 79 |
+
|
| 80 |
+
💻 GitHub: [https://github.com/UCR-CISL/CooperScene](https://github.com/UCR-CISL/CooperScene)
|
| 81 |
+
|
| 82 |
+
🤗 Hugging Face: [https://huggingface.co/cisl-hf/CooperScene](https://huggingface.co/cisl-hf/CooperScene)
|