Display markers, entities, camera images, meshes, URDF models, and more in a 3D scene.

Supported messages

The 3D panel can visualize an assortment of different messages.

To visualize a topic, the messages on that topic must conform to one of the known message schemas listed below.

Camera field-of-view

Calibration parameters for your scene's camera.

framework	schema
ROS 1	sensor_msgs/CameraInfo
ROS 2	sensor_msgs/msg/CameraInfo
Custom	foxglove.CameraCalibration

Grid

2D colored grids.

framework	schema
ROS 1	nav_msgs/OccupancyGrid
ROS 2	nav_msgs/msg/OccupancyGrid
Custom	foxglove.Grid

foxglove.Grid settings

field	description
Color mode	One of: Flat: solid color Color map: pre-defined color palette Gradient: smooth transition between two custom colors RGBA (separate fields): use embedded color from each cell's red, green, blue, and alpha fields (see below)
Flat color	Only shown if "Color mode" is set to "Flat"; hex code for color of each cell
Color by	Only shown if "Color mode" is not set to "Flat"; numeric field in message used for coloring logic
Color map	Only shown if "Color mode" is set to "Color map"; "Turbo" (Google) or "Rainbow" (RViz); for mapping "Color by" field values to colors
Opacity	Only shown if "Color mode" is set to "Color map" or "BGR (packed)"; sets alpha value for all cells
Value min	Only shown if "Color mode" is not set to "Flat"; minimum value used to normalize incoming grid's "Color by" field values
Value max	Only shown if "Color mode" is not set to "Flat"; maximum value used to normalize incoming grid's "Color by" field values
Frame lock	"On" means the grid is locked to the frame specified by its frame_id, and will move as that frame's transforms change. "Off" means the grid is relative to the fixed frame and will not move after it is first displayed.

RGBA (separate fields) color mode

Each cell can contain color information in four separate fields, named red, green, blue, and alpha, of any numeric type:

• Floating-point values — 0–1 range
• Unsigned integer values — Maximum possible range (e.g. 0–255 for a UINT8 field)
• Signed integer values — -max to max (e.g. −127 to 127 for an INT8 field; a value of −128 is treated as identical to −127)

nav_msgs/OccupancyGrid settings

field	description
Color mode	One of: Costmap: pre-defined RViz color palette. Cannot customize settings further. Custom: custom color palette using settings below
Min color	Color corresponding to minimum cell value (0)
Max color	Color corresponding to maximum cell value (100). Note that cells with value exactly 100 are displayed as fully transparent.
Unknown color	Color corresponding to unknown cell value (−1)
Fallback color	Color corresponding to cell values that fall outside the range from −1 to 100
Frame lock	"On" means the grid is locked to the frame specified by its frame_id, and will move as that frame's transforms change. "Off" means the grid is relative to the fixed frame and will not move after it is first displayed.

Image

Images displayed in the 3D scene, using the corresponding Camera field-of-view messages.

framework	schema
ROS 1	sensor_msgs/Image
ROS 2	sensor_msgs/msg/Image
ROS 1	sensor_msgs/CompressedImage
ROS 2	sensor_msgs/msg/CompressedImage
Custom	foxglove.RawImage
Custom	foxglove.CompressedImage

Laser scan

A single scan from a planar laser range-finder.

framework	schema
ROS 1	sensor_msgs/LaserScan
ROS 2	sensor_msgs/msg/LaserScan
Custom	foxglove.LaserScan

ROS polygons

Timestamped polygons made up of a series of connected points.

framework	schema
ROS 1	geometry_msgs/PolygonStamped
ROS 2	geometry_msgs/msg/PolygonStamped

ROS markers

Similar to scene entities, these Marker messages describe primitive shapes or meshes.

framework	schema
ROS 1	visualization_msgs/Marker
ROS 2	visualization_msgs/msg/Marker
ROS 1	visualization_msgs/MarkerArray
ROS 2	visualization_msgs/msg/MarkerArray

Mesh markers

Markers with a mesh_resource field support the following URL schemes:

• http(s)://
• package:// (Desktop app only)
• file:// (Desktop app only)

And file formats:

• glTF
• STL
• COLLADA
• Wavefront OBJ

glTF (.glb)

This is the preferred format, as it enjoys the best performance of all supported file types.

Binary glTF files bundle all required assets into a single file, with support for embedded meshes, compression, and the same physically-based material system used in Flora. As a result, your model should appear in Flora similarly to how it appears in other 3D programs.

STL (.stl)

STL files are well supported in Flora, but lack some of glTF's visualization features. The main advantage to STL is the ability to share the same files between your hardware manufacturing process and robot visualization tooling.

STL was designed for 3D printing and CAD applications, and does not include materials or hierarchies of meshes. While they can be represented in a binary encoding, STL files are commonly represented with ASCII characters, which leads to larger files.

COLLADA (.dae)

As a predecessor to glTF, COLLADA has a similar feature set. With that said, it does have larger XML-based files, no compression, and additional processing overhead.

There is a bug in RViz where the up-axis metadata is ignored, resulting in incorrect orientations for many .dae files in ROS environments. To work around this, the 3D panel has a Ignore COLLADA <up_axis> setting to toggle between observing the <up-axis> tag or ignoring it like RViz.

Wavefront OBJ (.obj)

OBJ is a simple ASCII format predating all other supported formats. It has large file sizes, no material support, no mesh hierarchies, no compression, and additional processing overhead.

Material support was added to the OBJ format as separate .mtl files, which Flora does not read.

Path

An array of timestamped poses in a named coordinate frame, denoting an object's path through space.

framework	schema
ROS 1	nav_msgs/Path
ROS 2	nav_msgs/msg/Path
Custom	foxglove.PosesInFrame

Point cloud

A collection of N-dimensional points, which may contain additional fields with information like normals, intensity, etc.

framework	schema
ROS 1	sensor_msgs/PointCloud2
ROS 2	sensor_msgs/msg/PointCloud2
Custom	foxglove.PointCloud

Settings

field	description
Point size	Size of each rendered point
Point shape	Shape of each rendered point
Decay time	Duration of time (in sec) that each point stays rendered
Color mode	One of: Flat: solid color Color map: pre-defined color palette Gradient: smooth transition between two custom colors BGR (packed): sensor_msgs/PointCloud2 only; use embedded color from each point's rgb field (see below) BGRA (packed): sensor_msgs/PointCloud2 only; use embedded color from each point's rgba field (see below) RGBA (separate fields): foxglove.PointCloud only; use embedded color from each point's red, green, blue, and alpha fields (see below)
Flat color	Only shown if "Color mode" is set to "Flat"; hex code for color of each point
Color by	Only shown if "Color mode" is not set to "Flat"; value used for "Color map" coloring logic; any numeric field in message such as x, y, z, <distance> (L2 norm of coordinates), or custom defined field
Color map	Only shown if "Color mode" is set to "Color map"; "Turbo" (Google) or "Rainbow" (RViz); for mapping "Color by" field values to colors
Opacity	Only shown if "Color mode" is set to "Color map" or "BGR (packed)"; sets alpha value for all points
Value min	Only shown if "Color mode" is not set to "Flat"; minimum value used to normalize incoming points' "Color by" field values
Value max	Only shown if "Color mode" is not set to "Flat"; maximum value used to normalize incoming points' "Color by" field values
Stixel view	Visualize points as stixels that extend from the point's z location to 0

RGBA color modes

When using the "BGR (packed)", "BGRA (packed)", and "RGBA (separate fields)" color modes, your point cloud message must contain certain fields to display color information for each point.

RGBA (separate fields)

For foxglove.PointCloud messages, each point can contain color information in four separate fields, named red, green, blue, and alpha, of any numeric type:

• Floating-point values — 0–1 range
• Unsigned integer values — Maximum possible range (e.g. 0–255 for a UINT8 field)
• Signed integer values — -max to max (e.g. −127 to 127 for an INT8 field; a value of −128 is treated as identical to −127)

BGR (packed) and BGRA (packed)

For sensor_msgs/PointCloud2 messages, each point can contain color information in a single field named rgb or rgba:

• Must use a 4-byte type from sensor_msgs/PointField (UINT32, value 6, is recommended)
• Each red, green, blue, and alpha value is represented by one byte in the 0–255 range
• Bytes must be packed in [0xBB, 0xGG, 0xRR, 0xAA] order (i.e. (0xAA << 24) | (0xRR << 16) | (0xGG << 8) | 0xBB in little-endian order). This order is compatible with RViz.

If using the "BGR" mode, the alpha value must still be present, but is ignored.

Pose

Poses in a named coordinate frame.

framework	schema
ROS 1	geometry_msgs/PoseArray
ROS 2	geometry_msgs/msg/PoseArray
ROS 1	geometry_msgs/PoseStamped
ROS 2	geometry_msgs/msg/PoseStamped
Custom	foxglove.PosesInFrame

Scene entity

A collection of primitive shapes (cubes, spheres, text, meshes, lines, etc) used to display anything from a basic bounding box to a complex 3D decision tree or road network.

Scene entities must be wrapped in a SceneUpdate message.

framework	schema
Custom	foxglove.SceneEntity
Custom	foxglove.SceneUpdate

Transforms

A transform (translation and rotation) between two reference frames in 3D space.

framework	schema
ROS 1	tf/tfMessage
ROS 1	tf2_msgs/TFMessage
ROS 2	tf2_msgs/msg/TFMessage
Custom	foxglove.FrameTransform

Velodyne scan

Velodyne Lidar scan packets from the Velodyne ROS driver or the Foxglove desktop app's Velodyne connection.

framework	schema
ROS 1	velodyne_msgs/VelodyneScan
ROS 2	velodyne_msgs/msg/VelodyneScan

Video

Compressed videos displayed in the 3D scene, using the corresponding Camera field-of-view messages.

framework	schema
Custom	foxglove.CompressedVideo

Settings

Frame

Select your 3D scene's display frame and follow mode.

field	description
Display frame	Configures the coordinate frame for rendering the scene. To render a scene element, you need a transform from its frame to the selected display frame. If its frame is the display frame, no additional transform is needed.
Follow mode	Configures the viewport following behavior. "Pose" - Follows the display frame's position and orientation "Position" - Follows the display frame's position only "Fixed" - Does not update viewport

Scene

Configure generic rendering properties and viewport properties.

field	description
Render stats	Display rendering performance statistics in panel corner
Background	Scene's background color
Label scale	Scale of rendered labels
Ignore COLLADA <up_axis>	Ignore the <up_axis> tag in COLLADA files
Mesh up axis	Direction of “up” when loading meshes (STL, OBJ) without orientation info ("Y-up", "Z-up")
Sync camera	Sync the camera states of multiple 3D panels

View

Configure the camera settings.

field	description
Distance	Distance of camera from viewport center (i.e. zoom level)
Perspective	Toggles between 3D ("on") and 2D ("off") view of the scene. The 2D view orients the camera towards z-down.
Target	Translational offset from the origin of the display frame
Theta	Azimuthal angle offset from the origin of the display frame in degrees
Phi	Polar angle offset from the origin of the display frame in degrees
Y-Axis FOV	Vertical field of view in degrees
Near	Near clipping plane distance
Far	Far clipping plane distance
Log depth	Enable logarithmic depth buffer for more uniform depth precision. May result in rendering artifacts or performance degradation.

Interaction of log depth and near/far plane settings

The Near, Far, and Log depth settings are interrelated. When using a very small Near value or a large Far value, enabling Log depth may reduce visual artifacts. For example, in the video below:

• The left panel does not use logarithmic depth, and it has a near clipping plane of 1.0. The grids render smoothly, but objects close to the viewer are clipped.
• The middle panel does not use logarithmic depth, and has a near plane of 0.01. The viewer can get closer to objects without clipping, but rendering artifacts (z-fighting) appear between multiple grids.
• The right panel enables the new logarithmic depth setting. This avoids z-fighting by providing more depth precision for the grids, while still allowing a near plane value of 0.01 for reduced clipping.

Transforms

A robotics system produces many messages describing its observations of the world around it. These messages may exist in one or more coordinate frames, or on different parts of the robot.

Transforms define the spacial relationships between two coordinate frames at a given time. To render objects into the scene, there must exist a transform path from the object's coordinate frame to the display frame.

Visualize transform frames to debug why elements may not be rendering where expected.

History

The panel stores coordinate frame relationships over time in a transform history store. Each coordinate frame stores up to 10,000 transform messages before it begins to remove the oldest messages as new messages arrive.

Preloading

Transform preloading in the 3D panel ensures an accurate scene by loading transform messages into memory. Transform preloading may impact 3D panel performance during preloading and seeking. You can disable preloading in the 3D panel settings. However, when disabled, the 3D panel may not properly render your scene in certain circumstances where infrequent coordinate frames are involved.

Troubleshooting

When working with live sources where time synchronicity cannot be achieved, it is recommended to have the server publish its Time according to the Foxglove Websocket Protocol. Otherwise if the server time is ahead of the visualization app, certain messages in the 3D panel will display immediately but coordinate frame state can lag behind as the app uses its wall time instead of the server time.

Settings

field	description
Editable	Toggles in-app editing of transform frames. When "on", you can update the Translation and Rotation values to tweak transform frames, see the impact on the scene, and debug how you may want to adjust the frames on your robot. When turned "Off", will revert to original translation and rotation values.
Labels	Toggle the display of transform labels.
Axis scale	Scale of transform axis (displayed as an arrow marker)
Line width	Width of transform (displayed as a line marker)
Line color	Color of transform (displayed as a line marker)
Enable preloading	Preload transform messages from the data source.

Topics

Lists topics containing supported message types that can be displayed in the scene. Toggle individual topics.

Each topic provides settings to configure how the visualization appears in the scene.

Any topic with messages matching the supported schemas will show up in the topics list. Each supported format has format specific visualization settings. These settings will show up under the topic.

Use the top-level menu to toggle all topics. You can toggle any individual topic by click the eye icon next to the topic name.

Custom layers

Custom layers allow you to add visual elements not from your data source.

Use the top-level menu to add custom layers.

Add Grid

A grid is 2D square with a fixed size and number of divisions. You can create any numbers of grids. Grids can be rendered relative to the display frame or any other transform frame.

Use the grid layer settings to change the frame, size, color, divisions, and other properties of the grid.

Add URDF

URDF robot models are loaded automatically if your data source supports parameters (i.e. a native ROS 1 or ROS 2 connection) and the /robot_description parameter is set to valid URDF XML.

You can add URDF models with a custom layer. Use each layer's menu to duplicate or delete the custom layer.

field	description
Source	Source of the URDF URL – URL to the source URDF file in one of the following schemes: http(s):// – Over an HTTP(S) connection package:// – Package path. Only supported when using the desktop app or a live connection that supports asset fetching, such as foxglove_bridge. File path – Absolute file path to the source URDF file. Only supported when using the desktop app. Parameteter – Parameter containing the source URDF file. Topic – Topic containing the source URDF file.
Label	(Optional) Label with which the custom layer will appear in the sidebar.
Frame prefix	(Optional) Prefix for robot's transforms. Also commonly known as a TF prefix.
Display mode	Robot link geometries to display Auto – Defaults to visual geometries, but falls back to collision geometries if there are no visuals Visual – Visual geometries Collision – Collision geometries
Color	Fallback color for the URDF model, if the source doesn't contain colors

Resolution of URDF assets with package:// URLs

Foxglove is able fetch meshes and other assests referenced by package:// in a number of different ways:

1. When using the desktop app, package paths can be resolved by finding packages in directories specified by the ROS_PACKAGE_PATH setting
2. When using a live connection that supports asset fetching such as Foxglove Bridge, assets can be fetched over the websocket connection
3. When the URDF is given as http(s) or local file path, package paths can be resolved when the URL or file path contains the package name. Consider the following example where the package name leo_description is contained in the URDF URL:
   • URDF URL: https://raw.githubusercontent.com/LeoRover/leo_common-ros2/humble/leo_description/urdf/leo.urdf.xacro
   • Asset URL: package://leo_description/models/Rocker.dae
   • Resolved asset URL: https://raw.githubusercontent.com/LeoRover/leo_common-ros2/humble/leo_description/urdf/leo.urdf.xacro

Publish

Configure click-to-publish behavior for the 3D panel.

field	description
Type	Type of message to publish. Point Pose Pose estimates
Topic	Topic to publish message to – only possible with a ROS data source with publish support.

User interactions

Click any object in the scene to display its relevant details in the selected object popup.

Click the icons next to the topic to open its settings in the sidebar or to display its messages in a new Raw Messages panel.

Selecting an object will set a $selected_id variable to the clicked object's id value, if it has one. For ROS markers and scene entities, you can also set another variable on click.

The panel controls on the right can be used to do the following:

• Select toggle object select mode
• 3D Toggle between 3D and 2D views of the scene
• Measure measure the distance between two points

Shortcuts

To move the camera:

• w – Move camera forward
• s – Move camera backward
• a – Move camera to the left
• d – Move camera to the right
• Scroll – Zoom in and out
• Drag – Move camera parallel to the ground
• Right-click + Drag horizontally – Rotate the camera around the world's z-axis
• Right-click + Drag vertically – Pan the camera around the world's x- and y- axes

Performance troubleshooting

If you are experiencing poor performance while interacting with the 3D panel, it may be because hardware acceleration is not enabled. Hardware acceleration will significantly improve performance by rendering on the GPU rather than the CPU.

If you're using Google Chrome, you can check that hardware graphics acceleration is enabled by going to chrome://gpu and seeing that "WebGL" and/or "WebGL2" are "Hardware accelerated". If not, go to chrome://settings, search your settings for acceleration and ensure that the "Use graphics acceleration when available" toggle is enabled. If after these steps you still see that WebGL is still not "Hardware accelerated" in chrome://gpu then you might have to investigate platform-specific steps for enabling your GPU.

The desktop app uses Electron (based on Chromium), and should automatically use hardware acceleration when available. If you're seeing issues here, we recommend following the steps above to see if the same issue exists in Chrome. If so, there's a good chance it's affecting our application for the same reasons. Therefore we recommend getting things working in Chrome first, and if the desktop app is still having issues afterward please let us know and we'll look into it.

Some reasons hardware acceleration might be disabled:

• Your graphics drivers are out of date or not installed such that Chrome can make use of them.
• Your system configuration could be unsupported for hardware acceleration in Chrome. A way to solve this is to enable "Override software rendering list" in about://flags, which can allow Chrome to use hardware acceleration on unsupported systems. This can result in an unstable Chrome experience and is not advised unless necessary.

If you are experiencing poor performance with the 3D panel while using foxglove.CompressedVideo topics, check that video acceleration is enabled. Docs here.

Links and resources

• Visualize point clouds with custom colors
• Visualize ROS mesh markers
• Write a message converter extension (3D panel)