In V4Design, visual services can recognise the type of a building depicted in a video scene and also localise this building, along with some construction (e.g. road, pavement, fence) and building facade elements (e.g. windows and doors). In the demos , we can see the masks that our algorithm extracted:
In red colour we see the pixels that are marked as building, in blue the windows, in pink the road and in grey the fence localised.
There is also info provided about the confidence of building existence in the video scene (building probability), the recognised type of the building (scene recognised) and the corresponding confidence level, as well as all the elements detected in the video frames (tags).