How to track objects in the real world with TensorFlow, SORT and OpenCV

Many of the solutions we work on at HAL24K involve detecting, classifying and tracking objects from camera feeds. However, for some of our customers it's not enough to simply identify and track each object in the frame: they need to know exactly where it is located in the real world (i.e. its latitude and longitude), how large it is or how fast it is travelling. Luckily, two factors come into play which we can use to our advantage on many of our projects. Firstly, the cameras we use are often fixed in position. This means that each pixel of the background always corresponds to the same real world location.