---
title: co_rec_track2
canonical_url: "https://www.modelscope.cn/datasets/eco210923/co_rec_track2"
md_url: "https://www.modelscope.cn/datasets/eco210923/co_rec_track2.md"
repository: eco210923/co_rec_track2
chinese_name: "红枫2026赛道2目标检测数据集"
last_updated: 2026-08-08
license: Apache-2.0
storage_size: "920 MB"
downloads: 59921
stars: 0
---

# co_rec_track2

> co_rec_track2 - eco210923 在 ModelScope 开源的数据集。红枫2026 赛道2 目标检测模型（真目标 / 诱饵）

eco210923/co_rec_track2 是 ModelScope 魔搭社区上的数据集，存储大小 920 MB，采用 Apache-2.0 许可。

- **Repository**: eco210923/co_rec_track2
- **License**: Apache-2.0
- **Storage size**: 920 MB
- **Downloads**: 59921
- **Stars**: 0
- **Last updated**: 2026-08-08

Source: https://www.modelscope.cn/datasets/eco210923/co_rec_track2

---

# 红枫2026 赛道2 目标检测模型（真目标 / 诱饵）

本仓库提供赛道2 协同算法的**视觉识别模型权重** `weights/best.pt`（YOLOv8n，imgsz=1024，~6MB）。

## 内容

| 文件 | 说明 |
|------|------|
| `weights/best.pt` | YOLO 视觉识别模型：区分真目标（TargetVehicle）与诱饵（DecoyVehicle）|

## 用法

1. clone 算法仓库：`git clone https://github.com/OpenHUTB/co_rec.git`
2. 下载模型：`modelscope download --repo-type dataset eco210923/co_rec_track2 --local-dir .`
3. 设置模型路径（**必须**，否则程序启动会崩溃）：
   ```bat
   setx YOLO_MODEL "C:/你的路径/weights/best.pt"
   ```
4. 把算法 `algorithm/coop_improved.py` 放到官方环境 `competition/user_algorithms/coop_decoy/`，启动 coop_decoy 场景。

完整步骤见算法仓库 `docs/for-collaborators.md`。
