Mirror of wlc952/FunASR-bmodel from ModelScope. All credit to the original authors. Not an official release from them. No license was declared at the source; all rights remain with the original authors.

FunASR-bmodel for monitor

快速开始

git lfs install
git clone https://www.modelscope.cn/wlc952/FunASR-bmodel.git

cd FunASR-bmodel
uv sync
source .venv/bin/activate

cd bmodel # 注意工作目录是 ./bmodel
python monitor.py   --rtsp rtsp://127.0.0.1:8554/stream1   --system-audio ../system_ref.wav  --system-phrases ../system_phrases.txt   --chunk-seconds 10

说明

  1. 本python项目使用uv管理,请先安装uv。参考安装方式:
# 推荐方式
curl -LsSf https://astral.sh/uv/install.sh | sh

# pipx 安装
sudo apt install pipx
pipx install uv
pipx ensurepath
  1. 虚拟环境创建/同步

    uv sync
    source .venv/bin/activate
    
  2. 关键文件

    • 模型测试脚本 bmodel/offline.py ,直接运行即可。
    cd bmodel
    python offline.py
    
    • 对话监测脚本 bmodel/monitor.py ,运行方式如下
    cd bmodel
    python monitor.py   --rtsp rtsp://172.24.64.225:8554/stream1   --system-audio ../system_ref.wav  --system-phrases ../system_phrases.txt   --chunk-seconds 10
    

    相关参数说明:

       parser = argparse.ArgumentParser(description="监控场景下非系统人声检测与上报")
       parser.add_argument("--rtsp", required=True, help="RTSP 地址,例如 rtsp://user:pass@ip:554/Streaming/Channels/101")
       parser.add_argument(
           "--system-audio",
           default="../system_ref.wav",
           help="系统参考音频,作为前置拼接,系统说话人 spk=0",
       )
       parser.add_argument("--out-dir", default="/data/alarms", help="告警本地保存目录 (已由 nginx 映射到 /static)")
       parser.add_argument("--chunk-seconds", type=int, default=10, help="检测窗口秒数,默认 10s")
       parser.add_argument(
           "--system-phrases",
           default="../system_phrases.txt",
           help="系统提示词文本库(每行一条),用于过滤系统播报文本。如果系统声音错误上报,请将对应的识别文本加入到该文件。",
       )
    

    还有一些参数需要直接修改代码:

    • ip修改:搜索并修改get_local_ip_fallback()
    • POST信息修改:搜索并修改def post_alert(safety_url: str, when: str, timeout: int = 2)
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including ml-resources/FunASR-bmodel