Using the Wizart Vision Detection API, you should pay attention that the data can be returned in two different forms, and it is more likely to use them for different needs
Bounding box is an imaginary rectangle drawn around the object that describes its spatial location. It’s the most basic tool used for object detection and localization tasks so you can determine position and as an option count objects of some class
Some surface detection tasks, such as wall detection, floor detection, and even window and door detection, work differently. This method is based on key points. Using these points, you can still localize the object by joined the points and thereby obtaining a frame around the detected object. Since the resulting border is not rectangular one, it is possible to describe more complex shapes, even taking into account the perspective, and determine more correct values for the object's coordinates. Also, this method allows you to get the coordinates of the invisible parts of the object, overlapped by other objects in the scene. You can see it on the bottom picture pay attention that some floor coordinates are hidden and overlapped by sofa and by the door.
see documentation
[
{
"wall_id": 0,
"points": [
[
0.0,
-0.32415005564689636
],
[
0.49296873807907104,
0.1488862782716751
],
[
0.0,
-0.32415005564689636
],
[
0.49296873807907104,
0.14888626337051392
]
],
"cx": 18.169554471969604,
"cy": 29.29260683227396
},
{
"wall_id": 1,
"points": [
[
0.49296873807907104,
0.1488862782716751
],
[
0.918749988079071,
0.03399765491485596
],
[
0.49296873807907104,
0.14888626337051392
],
[
0.918749988079071,
0.03399762511253357
]
],
"cx": 88.99655342102051,
"cy": 33.05871209952924
},
{
"wall_id": 2,
"points": [
[
0.918749988079071,
0.03399765491485596
],
[
1.0,
-0.43728017807006836
],
[
0.918749988079071,
0.03399762511253357
],
[
1.0,
-0.43728017807006836
]
],
"cx": 95.99307060241699,
"cy": 33.63378419686035
}
]