Method of estimating a three-dimensional position of an object

A method of estimating a three-dimensional position of at least one target object performed by a computer, the method comprising:

• obtaining a first image depicting a scene captured by a physical camera, wherein the scene comprises the at least one target object,
• detecting at least one target object, wherein the at least one target object is detected using an object detector on the first image, wherein detecting comprises identifying, by the object detector, the at least one target object and generating, by the object detector, coordinates indicating a position of the at least one detected target object,
• generating a set of images depicting subsets of the captured scene, the subsets of the scene being depicted by at least one virtual camera,
• estimating parameters (A, a, μ) of a probability distribution (P) indicative of a likelihood that the at least one target object is in a particular position in the scene by providing the generated set of images to a trained model, wherein the probability distribution P defines a likelihood P(x, y, z) that the target object is located at a particular position (x, y, z) in the scene, and wherein the probability distribution P is defined such that its density is proportional to
• exp
• ⁡
• (
• -
• f
• ⁡
• (
• 
• A
• ⁡
• (
• x
• z
• y
• z
• -
• μ
• x
• μ
• y
• )
• 
• ⁢
• z
• μ
• z
• )
• )
• ⁢
• μ
• z
• z
• ⁢
• exp
• ⁡
• (
• a
• ·
• g
• ⁡
• (
• log
• ⁡
• (
• z
• μ
• z
• )
• )
• )
• where A is a 2×2 matrix, a is a scalar parameter related to depth standard deviation, and μ=(μx, μy, μ_z) is a vector representing an average position, and
• estimating a three-dimensional position of at least one target object using the probability distribution (P).