Speed Up Your Python Code With Broadcasting and PyTorch

Back when I did my masters thesis I spent a lot of time processing large amounts of lidar data. One of these steps was to remove all point measurements that belonged to static objects in the scene (buildings, fences and so on). Each static object in the scene was modeled as a rectangular object, which essentially meant that I had to check if each lidar measurement fell inside any of the rectangles. The lidar that was used in my thesis operated at 10Hz, and each scan contained around 100,000 to 150,000 measurements, which meant that one second of lidar data corresponded to 1-1.5 million lidar points that had to be processed. At that time I did not know about Python or broadcasting, so my implementation of this processing step was not that fast or efficient.