Camera model of the scene (bottom) and measurements in image space (top).

Contexts in source publication

Context 1

... start by describing the image formation model that allows us to associate 3D camera parameters, 3D object sizes (i.e. heights) and 2D bounding boxes. This is also illustrated in Fig. ...

Context 2

... assume the world is composed of a dominant ground plane on which all objects are situated, and a camera that observes the scene. We adopt a perspective camera model similar to [16,17], which is parameterized by camera angles (yaw ϕ, pitch θ and roll ψ), focal length f and camera height h cam to the ground (see Fig. 2). For the measurements in the vertical axis of image frame, the location of the horizon is v 0 , while the vertical image center is at v c . Each object bounding box have a top v t and bottom v b location in the image. We assume all images were taken with zero roll, or were rectified beforehand [22]. We further assume, without loss of ...

Context 3

... start by describing the image formation model that allows us to associate 3D camera parameters, 3D object sizes (i.e. heights) and 2D bounding boxes. This is also illustrated in Fig. ...

Context 4

... assume the world is composed of a dominant ground plane on which all objects are situated, and a camera that observes the scene. We adopt a perspective camera model similar to [16,17], which is parameterized by camera angles (yaw ϕ, pitch θ and roll ψ), focal length f and camera height h cam to the ground (see Fig. 2). For the measurements in the vertical axis of image frame, the location of the horizon is v 0 , while the vertical image center is at v c . Each object bounding box have a top v t and bottom v b location in the image. We assume all images were taken with zero roll, or were rectified beforehand [22]. We further assume, without loss of ...