SENSUS Viewer

The Calyo SENSUS Viewer has been designed to run on any machine with a GPU. Best performance will be achieved with a dedicated graphics card, though this is unnecessary. The quality presets will allow you to tune the processing load to best suit your machine. Below is a table showing an indicative sample of optimal presets for some machines we tested on.

CPU	GPU	Operating System	Quality Preset	~FPS
Intel i9 12th Gen	NVIDIA 3080 Ti	Windows 10/11; Ubuntu 20.04	High	60
Intel i9 12th Gen	NVIDIA 3070	Windows 10/11; Ubuntu 20.04	Medium/High	60
M1	M1 GPU	OSX	Medium/High	60
Intel i7 11th Gen	Intel Iris X	Windows 10/11	Low/Medium	40
ARM Cortex-A57	NVIDIA Jetson Nano	NVIDIA Jetpack 4.5	Low/Medium	40

Supported Platforms
The app has cross-platform support for Linux, Windows and macOS. The following platforms have been tested and verified:

• Windows 10/11
• Linux: Ubuntu 20.04
• Mac: Sonoma

If your system is not found in the list above, it is likely the SENSUS Viewer will still run with no issues. However, we have not verified this ourselves, so we can offer no guarantees.

If you have not done so already, please follow the installation instructions for the SENSUS Viewer application before proceeding to the next section.

Using SENSUS Viewer #

Calyo SENSUS Viewer is an application that is available to evaluators to demonstrate this functionality in a convenient, visual, cross-platform package.

Upon opening the app, you will see a blank project page ready for loading in datasets or physical devices and running various algorithms, as shown below,

1. Global Options (PULSE OFF) #

These parameters are sent to the device on startup, so they must be selected before turning the sensor on.

Device Selection
The Device Selection dropdown allows various application modes and device selection.

Devices:
Once your device is plugged in you will be able to selected it from the dropdown. All Calyo PULSE serial numbers begin with the prefix “CP” e.g. (CP660EA810001).

⚠️ If you are unable to see your device please ensure it is plugged in correctly using a USB data and power cable. If issues persist please try unplugging the sensor and restarting the software before reaching out for additional support. Linux users please refer to the note below.

📌 Linux Note: For the sensor to be detected by the software you must first plugin the sensor and then run the, sudo rmod ftdi_sio command before launching the app. You will see an additional “Scan Devices” button which you must click before you will see the device in the device selection dropdown.

Dataset from file:
This option allows you to load and image a single raw signal data (XML) frame from a previously captured dataset. After selecting this option, a file section window will popup and prompt you to choose a raw dataset frame file.

Num Cycles:
The Num Cycles parameter defines the number of cycles in each transmission pulse.

This is loosely correlated to the output power of the sensor in transmission. Larger values may improve detections and the signal-to-noise ratio but result in an increase in the ‘thickness’ of detected objects and increased strength of multi-path reflections potentially reducing precision for some applications. Thus higher values are not always better, a balance is preferred, it’s recommended to test how this value affects your results, within the ranges shown below.

Max Distance (m)
The Max Distance parameter directly impacts the signal windowing and how far the sensor can detect. Smaller values are preferred as they increase the update rate of the sensor by reducing the wait time to receive echoes, so aim to set the value only as far as you need. The recommended maximum value is approximately **15 meters**.

2. Performance Options (PULSE OFF) #

Point Cloud Viewer > Performance Options
This window allows you to tune the performance of the app depending on your available computer hardware.

The performance pre-set will scale point-cloud and image resolution to an appropriate level (see System Requirements table above). In general, computers with dedicated GPUs will be able to provide higher performance, however it may be worth testing to see which pre-set works best for you.

3. Dataset Collection (PULSE OFF) #

Point Cloud Viewer > Performance Options
The application has the ability to collect data in the form of raw RF signals and point cloud data.
Before starting the sensor you can configure dataset collection by ticking the “Capture Signals” or “Capture Point Clouds” checkboxes (or both). You can name the dataset in the Prefix textbox.

• Signals: Captures the raw RF sensor data in a custom XML format. This format is useful as it allows you to tune the sensor settings after capture for further refinement.
• Point Clouds: This directly stores the point clouds in real-time that you see on your screen.

All files are stored in a new directory according to the prefix directly to the application folder.

4. Controller #

The controller window allows you to start, stop and pause the processing of the sensor data.
Please ensure you have setup the following settings before pressing play:

1. Global Options
2. Performance Options
3. (Optional) Dataset Collection Parameters

Once the previous parameters have been configured click play to start the sensor.

Once running you will be presented with a few new windows and options.

5. Global Options (PULSE ON) #

The global options window now displays many tuning parameters that you can adjust in real-time to configure the sensor for your application.

While imaging is running, you can adjust the level of noise present in the image by using the dB scale slider. Higher levels of noise might be preferable for inference tasks because they contain more detail, but they are harder to interpret.

The R Comp (short for radial compensation) slider can be used to compensate for distance. At a high level, signals in the far field will be brought up in intensity. It is akin to ‘focusing’ the sensor at a certain distance.

The Inner Radius slider can be used to set a deadzone (in metres) from the sensor. This is necessary because there is a direct transmission from the transmitter that will “drown out” the desired return signals if included in the image.

Filter Freq and Filter Bandwidth can be used to adjust the filter applied to the raw signals. To view the effect of these parameters, click on “view -> Spectrum Analyser” to see a frequency domain representation of each signal. A narrower bandwidth will be more effective at filtering out noise, however, no filter is perfect and too narrow a filter might cut out some of the desired signal too. If in doubt, the default value will be fine for most circumstances.

Lastly, Angle Limit can be used to focus the sensor to varying degrees in the forward direction. A lower angle limit will result in improved resolution in the forward direction but restricts the peripheral perception. Alternatively, the angle limiter can be turned off, useful if perception at wide (180 degree) FOV is required.

6. ROI Options #

The ROI options allow only a specific area of interest to be set for the 3-D point cloud imaging. This will ignore any readings outside of the configured area, useful if you only need readings from a specific volume in the sensor FOV. The units are in meters.

7. View Menu #

The View Menu allows you to access some additional views and graphs to help visualise or debug the sensor’s functionality and performance.

Waveforms
The waveforms view plots the raw or filtered data received on each channel, allowing you to view many or certain channels.

Spectra
The spectra plot shows the frequency response from a particular channel of the sensor. This should be around 40kHz, where a large peak should be observed, as shown below. Here, you can also see the filtering effect on the signals for fine-tuning.

[XZ] / [YZ] / [XY] Views
Lastly, 2-D plots can be visualised. (Z=Range, Y=Elevation, X=Width)

8. Structure #

The structure window displays various options to change the look and style of the 3-D point cloud. These changes will not be reflected in any exported data.