NCNN
@vionvision/camera-ui-ncnnv1.2.14
Add object and face recognition and number plate reading. Select a suitable configuration for your system.
- Version
- 1.2.14
- Updated
- October 1, 2026
- Requires
- ViON 2.2.5 or newer
- Licence
- MIT
Technical description
Detects objects, faces and license plates in the video of your cameras. It runs on the server's processor or, when one is present, on a graphics card that supports Vulkan.
What it does
- Detects people, vehicles and animals
- Detects faces and recognizes known people
- Detects license plates and reads their text
- Uses a graphics card with Vulkan when one is present, otherwise the processor
- Follows the confidence values from the camera's detection settings
- Downloads its models itself
What you need
- For faster detection, a graphics card with Vulkan support; without one the plugin runs on the processor
- For names on faces, this plugin must be selected as "Face Recognition" for each camera: camera settings, "Plugins", "Detections"
Settings
- Face recognition model: shared by all cameras. After a change the saved faces are recalculated
- Use Vulkan (GPU): when on, the models run on the graphics card through Vulkan if it is available; otherwise on the processor
- Vulkan device: which graphics card is used on systems with several; "Auto" leaves the choice to the system
- Active hardware: shows where the models are running
- Download models again: deletes the downloaded models and downloads the current ones
- Reset settings: puts all plugin settings back to their defaults
- Per camera, in the groups "Object detection", "Face detection" and "License plates": Model, Detection model and OCR model. "Default" follows the recommended model