@aijoey 用 MiniCPM-V 4.6 搭建了视觉智能体安全监控 demo：四个 CCTV 实时画面，模型观察近期帧窗口，判断活动是否正常，仅在识别到真实事件时调用 `raise_alert(reason, severity)`。模型持续将场景分类为“常规/无紧急”，只在高速公路出现警车蓝灯时触发警报，推理为“警车蓝灯亮起，可能表示紧急情况”。该演示展示了小 VLM 超越图像描述、实现实际智能体行为的潜力。
Thanks @aijoey for the impressive vision-agent security monitoring demo with MiniCPM-V 4.6.
What stood out to us is the model consistently classifying scenes as 'routine / no emergency'—and that's the essence of security AI. It's not about flagging every person or vehicle, but reading the situation and assessing urgency. Only when police car blue lights appear on the highway does it trigger an alert, with the reasoning: 'There are police car blue lights, indicating a potential emergency situation.'
This is exactly what we've been aiming for: vision that doesn't just see, but knows when to act.
Built a vision-agent security monitoring demo with MiniCPM-V 4.6. @OpenBMB Four CCTV style feeds run on a live clock. MiniCPM watches recent frame windows, decides whether activity is normal, and when it sees a genuine incident, it streams a tool call: raise_alert(reason, severity) The goal was to test whether a small VLM can move beyond captioning into practical agent behavior: observe pixels, reason about context, suppress normal activity, and trigger a structured action only when needed. What stood out: MiniCPM-V 4.6 handled the workflow well enough to feel useful for real world monitoring prototypes. Not just “what is in the image?”, but “should anything be done?” Small VLMs are getting very interesting for edge-ish and operational workflows.在 X 查看被引用的帖子
来源：OpenBMB · x.com