Virtual View: a simpler perspective on detected road objects

Open the right view

Select nawnaw or nawnaw pro in the AI model picker, then use the view selector to reach Virtual. These detector models provide Detection, Maps, Virtual and Parking in both plans. The segmentation model has a separate page, so switch back to a detector if you cannot find the four-view selector. Configure this before setting off.

What the simplified scene represents

Virtual View arranges supported detections in a simplified 2.5D scene. It removes much of the visual detail of the original camera image, making object categories easier to distinguish at a glance. The scene uses the same camera observations as Detection. Changing the viewpoint does not add a depth sensor or reveal hidden objects.

Read spacing and orientation as estimates

A drawn vehicle is a category representation. Its size, distance and direction may differ from the physical vehicle, especially when the view is partly blocked or the phone moves. A gap between rendered cars is not a measured clearance. Likewise, a stylized road or reference line does not prove that a lane marking was detected.

Compare the three perspectives

Detection is useful for seeing the source image with boxes. Virtual View emphasizes nearby detected categories in a simplified layout. 3D Maps adds mapped roads and buildings, navigation and weather; those buildings come from map data. Choose the presentation you find readable, and use direct observation of the road for driving decisions.

Check a setup while parked

With the phone mounted and the vehicle stationary, compare a visible object in Detection and Virtual View. Check whether its category and general side agree. This is a practical setup check, not proof of accuracy at every distance. If the scene looks confusing, return to the camera view; do not spend a journey trying to interpret uncertain geometry.