ML Video Detection / Classification
Introduction
This page contains setup instructions for enabling ML Video Detection / Classification in Witness™. These instructions apply to both ClearWay™ and AdvanceGuard®.
Contents
Licensing ML Object Detection
ML Video Object Detection requires a valid licence to run. This is obtained by using the https://navtechradar.atlassian.net/wiki/spaces/TUN/pages/100564993.
The following license options included in the licensing tool are for various ML algorithms; the latest is the highlighted option ML Object Detection V11 Plugin. Select this option, unless you are explicitly trying to run a depreciated version of the ML video detection.
The license will be added to the licensing tool, license options as follows:
Follow the rest of the instructions covered here:https://navtechradar.atlassian.net/wiki/spaces/TUN/pages/100564993to complete the license, and to apply it to either ClearWay™ or AdvanceGuard®.
Copying the ML Object Detection Model
You should install the appropriate ML model into the folder C:\ProgramData\Navtech\ML for it to function correctly. The latest model, which matches the plugin above, is Yolo11_n_model.onnx. Alternatively, if you are using the Video Analytics Tool: https://navtechradar.atlassian.net/wiki/x/AQC93Q then you should install this model; Yolo11_n_video_analytics_model.onnx.
To install a model, download it and copy it to the C:\ProgramData\Navtech\ML folder. Download links for all models can be found at the bottom of this page: ML Video Detection / Classification.
Configuring the Object Detection Model in Witness™
Follow the instructions for Adding a Camera: https://navtechradar.atlassian.net/wiki/x/24AWLw.
Then, follow the instructions for Assigning a Camera: https://navtechradar.atlassian.net/wiki/x/HwEXLw.
Next, create an Image Server Camera Config for the camera that you just added as documented here: Image Servers.
Finally, assign the camera config to the image server: https://navtechradar.atlassian.net/wiki/x/NYAYiw.
Verifying the Object Detection Model is Working
As the default Image Server Camera Config setup also sets up a re-stream of the camera video, overlayed with the ML object detections, you can view this in a program such as VLC to verify the ML is working correctly.
Open VLC, select Media, and then Open Network Stream…
Type the following URL into the network URL box: rtsp://127.0.0.1:8090 (change this to match the port if you have configured a different port). In addition, if you are connecting to an image server on a different machine, then change the IP address to that of the machine you are connecting to.
If all is setup correctly, you will see the camera image being streamed back out from the image server and into VLC. You will also see detections (bounding boxes and labels) appearing around certain objects in the scene as in the following: