apps/annotator/ 使用纯 HTML、CSS 和 JavaScript,不需要前端构建。它必须由本地 Python 后端提供服务,直接打开 file:// 页面无法调用 detector。
python -m pip install -e ".[dev]"
python -m hotarea_cv.server \
--annotator-dir apps/annotator \
--dataset-root ./data \
--open-browser默认监听 127.0.0.1:8765。Windows 用户也可双击根目录的 启动 Jingwei 经纬标注平台.cmd;旧文件名仅作为兼容入口保留,原开发电脑上的桌面快捷方式不是可移植发布物。
上传单图最大 50 MB。Detector 运行结果先保存在后端内存,默认 600 秒后过期;执行导出后写入:
<dataset-root>/candidates/<image_id>.candidates.json
<dataset-root>/overlays/<image_id>.overlay.png
<dataset-root>/reports/<image_id>.report.json
Report 使用 run_id 和 SHA-256 绑定 candidates 与 overlay。服务只加载本仓库 OpenCV detector,不加载、下载或分发 best.pt 权重。默认服务是本地工具,不是经过加固的公网部署。
apps/annotator/ uses plain HTML, CSS, and JavaScript with no frontend build. It must be served by the local Python backend; a file:// page cannot call the detector.
python -m pip install -e ".[dev]"
python -m hotarea_cv.server \
--annotator-dir apps/annotator \
--dataset-root ./data \
--open-browserThe default listener is 127.0.0.1:8765. Windows users may run 启动 Jingwei 经纬标注平台.cmd; the old filename remains only as a compatibility entry point, and a desktop shortcut from the original development machine is not a portable release artifact.
One uploaded image may be at most 50 MB. Detector runs stay in backend memory and expire after 600 seconds by default. Export writes:
<dataset-root>/candidates/<image_id>.candidates.json
<dataset-root>/overlays/<image_id>.overlay.png
<dataset-root>/reports/<image_id>.report.json
The report binds candidates and overlays with a run_id and SHA-256 values. The service loads only this repository's OpenCV detector; it does not load, download, or distribute a best.pt weight. The default server is a local tool, not a hardened public deployment.