Run examples

Run them from <prefix>\example\build\Release on Windows, or <prefix>/example/build on Ubuntu.

The four viewers

Every viewer takes --config=<prefix>/config and starts in color unless you pass color_depth or pointcloud.

View controls are the same in all four. Switch the view live with 1 / 2 / 3 - colour, colour + depth, point cloud. In point-cloud mode, drag to orbit and scroll the wheel to zoom. ESC quits.

[!NOTE] A complete multi-camera frame group only forms when the camera hosts are PTP-synced. A single camera always forms complete groups; a second one joins once the two hosts share a clock. See Set up time sync (PTP).

viewer_opengl_streaming

Live point clouds from a fixed device list.

viewer_opengl_streaming pointcloud --devices=<id-1>,<id-2> --config=<prefix>

viewer_opengl_streaming pointcloud --devices=<id-1>,<id-2> --config=<prefix>

viewer_opengl_streaming pointcloud --devices=<id-1>,<id-2> --config=<prefix>

Or with a JSON device list:

viewer_opengl_streaming pointcloud --clients=<prefix>/example/streaming_client_info_example.json --config=<prefix>

viewer_opengl_streaming pointcloud --clients=<prefix>/example/streaming_client_info_example.json --config=<prefix>

viewer_opengl_streaming pointcloud --clients=<prefix>/example/streaming_client_info_example.json --config=<prefix>

Press R to record each camera (files video_<device-id>_<n>.mkv), S to stop.

The --clients file

A small JSON listing the device IDs to stream - the same set you would pass to --devices, in a file. Handy for a fixed rig or many cameras. A ready-to-copy template lives at <prefix>/example/streaming_client_info_example.json.

{
  "client_info": [
    { "client_id": "AAAA-BBBB-CCCC-0129" },
    { "client_id": "AAAA-BBBB-CCCC-0138" }
  ],
  "client_num": 2
}

{
  "client_info": [
    { "client_id": "AAAA-BBBB-CCCC-0129" },
    { "client_id": "AAAA-BBBB-CCCC-0138" }
  ],
  "client_num": 2
}

{
  "client_info": [
    { "client_id": "AAAA-BBBB-CCCC-0129" },
    { "client_id": "AAAA-BBBB-CCCC-0138" }
  ],
  "client_num": 2
}

To fill it in for your own rig:

1. List your real device IDs - launch viewer_opengl_with_manager and press I.

2. Add one client_info entry per camera. Every ID must have a matching config/<device-id>/ folder under --config.

3. Set client_num to the entry count. Two entries means "client_num": 2.

--clients is mutually exclusive with --devices.

viewer_opengl_loader

Plays back recorded .mkv files. No cameras needed.

1. Get the recordings. They come from the R key in the streaming, manager or segment viewers.

2. Write a --videos JSON pairing each client_id with its video_file. Use full paths with forward slashes. The client_id must match the recorded device so the loader can read that device's config/<client_id>/ for intrinsics and extrinsics.

   {
     "client_num": 2,
     "client_info": [
       { "client_id": "AAAA-BBBB-CCCC-0129", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0129_0.mkv" },
       { "client_id": "AAAA-BBBB-CCCC-0138", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0138_0.mkv" }
     ]
   }

   {
     "client_num": 2,
     "client_info": [
       { "client_id": "AAAA-BBBB-CCCC-0129", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0129_0.mkv" },
       { "client_id": "AAAA-BBBB-CCCC-0138", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0138_0.mkv" }
     ]
   }

   {
     "client_num": 2,
     "client_info": [
       { "client_id": "AAAA-BBBB-CCCC-0129", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0129_0.mkv" },
       { "client_id": "AAAA-BBBB-CCCC-0138", "video_file": "E:/recordings/video_AAAA-BBBB-CCCC-0138_0.mkv" }
     ]
   }

3. Run it:

   viewer_opengl_loader color_depth --videos=my_videos.json --config=<prefix>

   viewer_opengl_loader color_depth --videos=my_videos.json --config=<prefix>

   viewer_opengl_loader color_depth --videos=my_videos.json --config=<prefix>

Playback keys: P pause, R resume, S reset to 2 s, T print time. Multiple recordings are auto-aligned and synchronized on playback.

viewer_opengl_with_manager

Discovers and connects cameras at runtime. It takes no --devices or --clients - it finds cameras over mDNS. It still takes --config.

viewer_opengl_with_manager --config=<prefix>

viewer_opengl_with_manager --config=<prefix>

viewer_opengl_with_manager --config=<prefix>

Drive it from the keyboard in the viewer window. Keys marked (asks for id) prompt you to type a device ID in the terminal and press Enter.

Typical run: launch → I to see the cameras → X (or C per ID) to connect → L each camera ID to start streaming → view the fused cloud. Each streaming camera attaches to and detaches from the synchronizer automatically.

Recording with R saves on the server side, under the server's videos/ folder as Recording_<device-id>_<n>.mkv for every connected camera. S stops. Client-side recording on the device is not used.

viewer_opengl_segment

Streaming plus 3D segmentation. Takes the same fixed-device-list arguments as streaming.

viewer_opengl_segment --devices=<id-1>,<id-2> --config=<prefix>

viewer_opengl_segment --devices=<id-1>,<id-2> --config=<prefix>

viewer_opengl_segment --devices=<id-1>,<id-2> --config=<prefix>

G segments once; H toggles continuous.

[!IMPORTANT] Segmentation requires the Python detector service. Set it up first - Set up the detection system. Without it the viewer still streams the live cloud but reports segmentation unavailable.

Command-line reference

Arguments may appear in any order. A bare word that is not a render mode and does not start with -- is treated as the config path.

[!NOTE] viewer_opengl_with_manager takes none of --devices, --clients or --videos. It discovers cameras over mDNS and you connect and stream them with the keys above. It still takes --config.

Keyboard reference

Next steps

Set up the detection system for the segmentation overlay, or go straight to Use the SDK in your own project.