Skip to main content

​
Overview

VSLAM is a beta product. While we are actively refining and testing, you may encounter occasional bugs, performance dips, or unexpected behavior. Please submit bug reports and feature requests to our team at support@dwe.ai.
DeepWater Desktop can automatically create a dense reconstruction of your environment, useful for live coverage mapping and measurements.

​
Requirements

  • SVC Pro Module - Running dweOS Pro
  • Topside Computer/Laptop - Running DeepWater Desktop

​
Launch Parameters

1

Change network protocol

Open the “Add Stream” dialog, select your cameras, and switch the Network Protocol to Coverage Mapping. Check out our quickstart guide for info on how to do this.
2

Adjust tracking and filter parameters (Optional)

Stick to the default parameters for the best out-of-the-box experience. Manual tuning is entirely optional and reserved for advanced performance optimization.
VSLAM tracking performance can be optimized for your environment by adjusting the tracking and voxel filter parameters.

Tracking & Filter Parameters

Reference for tuning VSLAM tracking and voxel filter parameters.
3

Start the Stream

Hit the “Start Stream” button. Once the stream initiates, DeepWater Desktop will begin processing your data. You should now see a 3D point cloud being generated as your camera moves around.If you plan on post-processing the data, start a recording here as well.

​
Loop closures

All SLAM systems accumulate error over time. To correct for this, we use the concept of loop closures— viewing the same location from the same pose at different points in time. DeepWater Desktop automatically recognizes loop closures and adjusts previously measured poses and point clouds accordingly. To initiate a loop closure, simply travel back to a previous pose and face in the same direction. A message will pop up in the bottom-right corner:
indicating that a loop closure has happened, followed by a global adjustment of the pointcloud and previous poses.

Point cloud before loop closure with accumulated error


Point cloud after loop closure with corrected poses

​
Maps

Upon losing tracking for an extended period, the system will create a new map from scratch. All previous maps are saved in memory, and can be viewed by clicking their labels in the top-right corner:

The highlighted map is the one currently being added to.

Don’t worry about losing tracking over short periods, as maps are automatically merged when possible. Upon discovery of overlapping sections, you will receive this message:
and after a short wait, the point cloud will reflect the changes. Automatic map creation can be toggled on/off by pressing Z or the Map button. If it’s off the system will only try to locate itself relative to existing maps.

​
Tools and Controls

When the Scene view is focused, use the following controls to move the virtual camera:

Camera position can be reset by hitting this button.

​
Ruler

Hit Space to pause the stream. Press R or select the Ruler icon in the bottom-right corner to activate the Ruler tool. Select two points to see the distance between them.

​
Color Mode

Press C or select the color palette icon in the bottom-right corner to toggle between Texture and Coverage viewing modes.

Texture mode shows the average observed RGB color for each point.


Coverage mode shows the nearest distance at which each point was observed. Red is closer, blue is further. Uses the same color mapping as the depth view.

Coverage mode makes it easier to spot which sections are lacking detail.

​
Troubleshooting

​
Configuring UDP Receive Buffers

RTP packets can be discarded due to insufficient UDP receive buffer size. We recommend manually increasing it as the default is often insufficient for properly capturing full frames.
We recommend setting the max OS receive buffer size to 8MB: