📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SaaS · 58 分 ← 返回列表

Restaurants are using AI on their existing CCTV to track which tables are free or taken. I built an open-source version that alerts you on WhatsApp

I saw how some restaurants now watch every table in real time: green means free, red means customers are seated. The part that caught my attention is that it can run on the security cameras they already have. I built my own version solo for the Meta Global AI Developer Hackathon 2026. **What mine does today:** - You draw a zone over the tables with a pencil tool. Everything outside the line is ignored, so people walking past don't trigger anything. - Empty tables are shown

五维评分

付费意愿
55
需求真实度
40
个人可落地
85
竞争空间
50
新鲜度
70

AI 分析

痛点:餐厅无法实时知道哪张桌子空着、哪张已被占用、哪桌客人久等没人招呼,只能靠服务员来回巡视,翻台效率低。

目标用户:中小餐厅、咖啡馆的老板与门店运营经理;已有监控摄像头但不想增加硬件的中小连锁餐饮。

证据:"The part that caught my attention is that it can run on the security cameras they already have." / "When a guest actually sits down (the AI must be at least 75% sure), it sends a photo and the reasoning to WhatsApp." / "If you run a restaurant or cafe, which alert would help you most: a waiter-delay timer..."

变现:按门店或按摄像头的月度 SaaS 订阅(如每店 $29–99/月),核心检测开源、云端托管与多店看板付费;增值模块单独收费:久等未服务提醒、未清理桌提醒、高峰时段周报,以及 WhatsApp 消息量加价包。

MVP:2 周 MVP:如果已有开源代码,就在它的基础上改造——只支持 RTSP 拉流,用浏览器端轻量检测模型,用铅笔工具框选桌位区域;只实现两种告警:客人入座推送 WhatsApp 或 Telegram 消息,以及入座 5 分钟无人服务时二次提醒。提供一条 docker compose 命令,让商家自行部署到本地小主机;附上单店安装文档,也可加一个云托管测试版,验证付费意愿。

竞品:Density、V-Count、Presto (table management)、TabSquare、Senseforth/通用 CCTV 视频分析厂商、OpenTable/SevenRooms(桌位管理但不做视觉占用检测)

摘要:开发者用 Next.js、Meta Muse Spark(Baseten)、WhatsApp Cloud API 和浏览器端运动检测,独立做出一个开源方案:复用餐厅已有的 RTSP/HLS 或 YouTube Live 监控,用铅笔工具框选桌位区域;当 AI 置信度达到 75% 判定客人入座后,就把截图和推理结果推送到 WhatsApp,并用撞锁去重。作者正在征询下一步该做“久等未服务计时”还是其他告警,尚无收入或付费用户数据,也未提出任何商业化定价;需求目前只是作者对餐厅实时桌位的观察,热度数据为空。