docs/foxglove_visualization.md

# Foxglove Visualization

Language: English | [简体中文](foxglove_visualization.zh-CN.md)

This guide shows how to inspect the dataset in Foxglove with ToF heatmaps, RGB images, LiDAR point clouds, MAVROS IMU curves, and 3D odometry.

The raw rosbag remains the source data. The Foxglove bag is a visualization artifact generated with `scripts/foxglove_visual.py`.

## Ready-To-Open Example

The Hugging Face dataset includes a Foxglove example:

```text
examples/foxglove/visual_demo.bag
```

It is generated from session:

```text
raw/sessions/2026-05-20_030414_odom_run029/bag.bag
```

Open the example bag in Foxglove, then create the panels below.

## Recommended Layout

| Panel | Topic / setting |
| --- | --- |
| Image | `/foxglove/tof/overview/compressed` |
| Image | `/camera/color/image_raw` |
| 3D | `Fixed frame = odom`, `Display frame = odom` |
| 3D point cloud | `/foxglove/livox/points` |
| 3D path | `/foxglove/odom/path` |
| 3D odometry | `/fusion_odometry/lazy_point_odom` or `/ekf_quat/ekf_odom` |
| Plot | `/mavros/imu/data_raw.angular_velocity.{x,y,z}` |
| Plot | `/mavros/imu/data_raw.linear_acceleration.{x,y,z}` |

For this dataset, set both 3D frame fields to:

```text
Fixed frame:   odom
Display frame: odom
```

The visualization bag provides this TF chain:

```text
odom -> base_link -> livox_frame
base_link -> base_link_frd
odom -> odom_ned
```

Use `odom` for the main 3D view. `odom_ned` and `base_link_frd` are auxiliary frames for NED/FRD conventions.

## Topic Guide

| Topic | Type | What it shows |
| --- | --- | --- |
| `/foxglove/tof/overview/compressed` | `sensor_msgs/CompressedImage` | Six-node Nooploop TOFSense-M 8x8 cascade heatmap |
| `/nlink_tofsensem_cascade` | `nlink_parser/TofsenseMCascade` | Original ToF numeric data |
| `/camera/color/image_raw` | `sensor_msgs/Image` | Intel RealSense D435i RGB stream |
| `/camera/color/camera_info` | `sensor_msgs/CameraInfo` | RGB camera intrinsics |
| `/livox/lidar` | `livox_ros_driver2/CustomMsg` | Original Livox MID360s packet topic |
| `/foxglove/livox/points` | `sensor_msgs/PointCloud2` | Foxglove-ready LiDAR point cloud |
| `/foxglove/odom/path` | `nav_msgs/Path` | Accumulated odometry trajectory for 3D display |
| `/mavros/imu/data` | `sensor_msgs/Imu` | Filtered flight-controller IMU |
| `/mavros/imu/data_raw` | `sensor_msgs/Imu` | Raw flight-controller IMU |
| `/fusion_odometry/lazy_point_odom` | `nav_msgs/Odometry` | Main odometry for 3D trajectory |
| `/ekf_quat/ekf_odom` | `nav_msgs/Odometry` | EKF odometry, when present |
| `/tf` | `tf2_msgs/TFMessage` | Dynamic odometry transform |
| `/tf_static` | `tf2_msgs/TFMessage` | Static camera, LiDAR, and frame transforms |

Foxglove does not reliably render Livox custom messages directly. Use `/foxglove/livox/points` for the 3D point cloud.

## ToF Panel

Add an Image panel and select:

```text
/foxglove/tof/overview/compressed
```

The overview image contains six TOFSense-M nodes. Each node shows:

| Field | Meaning |
| --- | --- |
| `dis` | distance in millimeters |
| `dis_status` | per-pixel distance status |
| `signal_strength` | per-pixel return strength |

Valid pixels are colorized by distance. Invalid or missing pixels are drawn in gray.

## LiDAR Panel

Add a 3D panel and select:

```text
/foxglove/livox/points
```

Recommended point-cloud settings:

| Setting | Value |
| --- | --- |
| Point size | `1.0` to `1.5` |
| Point shape | Circle |
| Color mode | Color map |
| Color field | `intensity` |
| Stixel view | Off |

If Stixel view is enabled, Foxglove draws pillar-like vertical structures. That is useful for obstacle-style views, but it is not the best mode for checking the raw point cloud.

## Odometry And TF

Add odometry display in the same 3D panel:

```text
/fusion_odometry/lazy_point_odom
```

For the already-traveled trajectory line, add:

```text
/foxglove/odom/path
```

If that topic is not available in a selected session, use:

```text
/ekf_quat/ekf_odom
```

The generated bag injects dynamic TF from odometry, so the 3D view can resolve:

```text
odom -> base_link -> livox_frame
```

This is the frame path required to show the LiDAR point cloud together with the odometry trajectory.

## IMU Plots

Add a Plot panel and select angular velocity:

```text
/mavros/imu/data_raw.angular_velocity.x
/mavros/imu/data_raw.angular_velocity.y
/mavros/imu/data_raw.angular_velocity.z
```

For acceleration, add:

```text
/mavros/imu/data_raw.linear_acceleration.x
/mavros/imu/data_raw.linear_acceleration.y
/mavros/imu/data_raw.linear_acceleration.z
```

Use `/mavros/imu/data` when you want the filtered MAVROS IMU stream, and `/mavros/imu/data_raw` when you want the raw flight-controller IMU stream.

## Generate A Visualization Bag

Build the ROS1 helper image once:

```bash
make docker-build
```

Generate a 30 second all-modality example:

```bash
docker/ros1_noetic/run.sh 'python3 scripts/foxglove_visual.py \
  --input-bag raw/sessions/2026-05-20_030414_odom_run029/bag.bag \
  --output-bag outputs_tmp/foxglove_samples/run029_mid30s_all_modalities_foxglove.bag \
  --start-offset-sec 189.891 \
  --duration-sec 30 \
  --copy-mode custom \
  --keep-topics /tf_static,/camera/color/image_raw,/camera/color/camera_info,/livox/lidar,/mavros/imu/data,/mavros/imu/data_raw,/nlink_tofsensem_cascade,/fusion_odometry/lazy_point_odom,/ekf_quat/ekf_odom \
  --tof-rate-hz 15 \
  --tf-rate-hz 30 \
  --tf-parent-frame odom \
  --force'
```

Generate a ToF + odometry focused bag:

```bash
docker/ros1_noetic/run.sh 'python3 scripts/foxglove_visual.py \
  --input-bag raw/sessions/2026-05-20_030414_odom_run029/bag.bag \
  --output-bag outputs_tmp/foxglove_samples/run029_tof_odom_foxglove.bag \
  --copy-mode custom \
  --keep-topics /tf_static,/mavros/imu/data,/mavros/imu/data_raw,/nlink_tofsensem_cascade,/fusion_odometry/lazy_point_odom,/ekf_quat/ekf_odom \
  --tof-rate-hz 15 \
  --tf-rate-hz 30 \
  --tf-parent-frame odom \
  --force'
```

## Useful Options

| Option | Purpose |
| --- | --- |
| `--start-offset-sec` | Start from an offset relative to the raw bag start |
| `--duration-sec` | Convert only a time window |
| `--copy-mode custom` | Keep exactly the original topics listed in `--keep-topics` |
| `--keep-topics` | Original raw topics to copy into the visualization bag |
| `--tof-rate-hz` | ToF overview image rate; `15` matches the nominal TOFSense-M 8x8 rate |
| `--tof-image-mode overview` | Write only the six-node overview image |
| `--tof-image-mode both` | Write overview plus per-node ToF images |
| `--convert-livox-pointcloud2` | Convert `/livox/lidar` to `/foxglove/livox/points` |
| `--livox-calibration` | LiDAR-to-body calibration YAML used for `/tf_static` |
| `--tf-parent-frame odom` | Use `odom -> base_link` for injected dynamic TF |
| `--bag-compression bz2` | Compress the generated bag |

## Troubleshooting

| Symptom | Check |
| --- | --- |
| ToF image panel says waiting for messages | Make sure `/foxglove/tof/overview/compressed` exists in the bag |
| LiDAR topic has a warning icon | Use `/foxglove/livox/points`, not raw `/livox/lidar` |
| Point cloud does not appear in 3D | Confirm `Fixed frame = odom` and `Display frame = odom` |
| Point cloud looks like vertical pillars | Turn Stixel view off |
| Odometry and point cloud are not aligned | Confirm `/tf` and `/tf_static` are enabled |
| RGB panel is blank | Use an Image panel for `/camera/color/image_raw` |

To inspect the generated bag before opening it:

```bash
docker/ros1_noetic/run.sh 'rosbag info outputs_tmp/foxglove_samples/run029_mid30s_all_modalities_foxglove.bag'
```