Automatically detecting the speed of an object in a video stream is a complex process that can require multiple models to detect the object, track it in subsequent frames, and based on this, estimate its speed. Existing speed detection models typically rely on multiple models for detecting and tracking moving objects. In this paper, we use a Multi-Task Learning (MTL) framework [5] to support an end-to-end deep learning model that detects and tracks vehicles (trucks) and estimates their speed. The proposed approach achieves an average speed estimation error of \(3.8\%\) , whilst providing faster inference than the state-of-the-art model [9].