Rdk X5 Ai Detect
v1.0.0在 RDK X5 的 10TOPS BPU 上运行单个 AI 推理算法：YOLO 目标检测、图像分类、语义分割、人脸识别、手势识别、人体关键点、开放词汇检测（DOSOD/YOLO-World）、双目深度估计、语音识别、端侧轻量 LLM（≤2B 参数量化模型）。Use when the user wants to...
0· 208·1 当前·1 累计
by 安全扫描
OpenClaw
安全
high confidenceInstructions are coherent with a ROS2-based RDK X5 BPU AI inference skill; the steps and requirements match the declared purpose, but take care with optional pip installs, example credential handling, and exposed web/RTSP streams.
评估建议
This skill appears internally consistent for running AI inference on an RDK X5 device. Before installing/using it: (1) confirm you are on the intended RDK X5 hardware with TROS/ROS2 installed; (2) avoid running the provided pip commands without verifying the packages' source (they may fetch code from the network); (3) do not embed credentials in command lines—use protected config files or environment variables and be aware such command lines can leak to process lists or shell history; (4) the we...详细分析 ▾
✓ 用途与能力
Name/description claim running AI inference on RDK X5 BPU via the TROS framework; SKILL.md only requires ros2 and references ROS launch files, BPU sysfs, and vendor Python inference packages—these are consistent with the stated purpose.
ℹ 指令范围
Runtime instructions are limited to sourcing the TROS environment, launching ROS2 nodes for specific AI components, optional pip installs, reading BPU usage from /sys, and viewing a local web UI. Nothing in the instructions asks for unrelated system-wide data, but the examples include an RTSP URL with credentials (exposes secrets on the command line) and a web UI on port 8000 (network exposure).
ℹ 安装机制
This is instruction-only (no install spec). It suggests optional pip installs (hobot-dnn-rdkx5, hobot-vio-rdkx5). Pip pulls from package indexes and may fetch remote code—verify package provenance before installing. No downloads from arbitrary URLs or archive extraction are prescribed.
✓ 凭证需求
No environment variables or external credentials are required by the skill. The only credential-like artifact is the RTSP example containing 'admin:password' which is an example only but could encourage insecure practices; otherwise no unexplained secret access is requested.
✓ 持久化与权限
always is false and there is no install that modifies global agent settings or other skills. The skill does not request permanent elevated privileges or persistent presence beyond normal invocation.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.02026/3/11
Initial release: BPU AI inference for YOLO, classification, segmentation, face/gesture/pose, DOSOD, depth, ASR, LLM
● 无害
安装命令 点击复制
官方npx clawhub@latest install rdk-x5-ai-detect
镜像加速npx clawhub@latest install rdk-x5-ai-detect --registry https://cn.clawhub-mirror.com
技能文档
RDK X5 搭载 10 TOPS BPU，所有 AI 算法通过 TogetheROS.Bot (tros.b) 框架运行。
前置准备
# 每个终端都需要 source
source /opt/tros/humble/setup.bash# 安装 Python 推理库（可选，用于自定义推理脚本）
pip3 install hobot-dnn-rdkx5 hobot-vio-rdkx5
操作步骤
1. YOLO 目标检测（摄像头实时）
source /opt/tros/humble/setup.bash
ros2 launch dnn_node_example dnn_node_example.launch.py \
dnn_example_config_file:=config/yolov5sworkconfig.json \
dnn_example_image_width:=640 dnn_example_image_height:=480
http://:8000 查看检测结果。可选模型配置文件：
| 模型 | config 文件 |
|---|---|
| YOLOv5s | config/yolov5sworkconfig.json |
| MobileNetV2 分类 | config/mobilenetv2workconfig.json |
2. 图像分类
source /opt/tros/humble/setup.bash
ros2 launch dnn_node_example dnn_node_example.launch.py \
dnn_example_config_file:=config/mobilenetv2workconfig.json
3. 开放词汇检测（DOSOD / YOLO-World）
source /opt/tros/humble/setup.bash# DOSOD（地瓜自研，支持语音指令指定检测目标）
ros2 launch hobot_dosod dosod.launch.py
# YOLO-World
ros2 launch hobot_yolo_world hobot_yolo_world.launch.py
4. 双目深度估计
source /opt/tros/humble/setup.bash
# 需配合双目摄像头（SC230ai / SC132gs / ZED）
ros2 launch hobot_stereonet stereonet_model.launch.py
5. 语音识别 (ASR)
source /opt/tros/humble/setup.bash
ros2 launch hobot_asr hobot_asr.launch.py
6. 端侧 LLM 大模型
source /opt/tros/humble/setup.bash
ros2 launch hobot_llamacpp hobot_llamacpp.launch.py
7. Web 可视化
source /opt/tros/humble/setup.bash
ros2 launch websocket websocket.launch.py
http://:8000 查看。8. RTSP 拉流 + AI 推理（智能盒子方案）
source /opt/tros/humble/setup.bash
ros2 launch hobot_rtsp_client hobot_rtsp_client.launch.py \
rtsp_url:="rtsp://admin:password@192.168.1.64:554/stream1"
性能监控
# BPU 使用率
cat /sys/devices/system/bpu/bpu0/ratio# 推理帧率
ros2 topic hz /ai_msg
排查故障
| 现象 | 原因 | 解决 |
|---|---|---|
Load model failed | 模型文件不存在或格式错误 | 检查 .bin 文件路径；模型需经 hb_mapper 转换为 NV12 格式 |
| BPU ratio 始终 0 | 推理节点未启动或模型未加载 | ros2 node list 确认 dnn_node 正在运行 |
| Web 无 AI 结果叠加 | websocket 未连接 AI topic | 确认 websocket 订阅的 topic 名称正确 |
| 帧率极低 (<5fps) | 模型过大或分辨率过高 | 降低输入分辨率；使用轻量模型（如 YOLOv5s 替代 YOLOv5x） |
数据来源：ClawHub ↗ · 中文优化：龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制
免费技能或插件可能存在安全风险，如需更匹配、更安全的方案，建议联系付费定制