Since the lockdown, we have been experiencing an increased interest in cycling and an increased number of pedestrians in city parks or nature. Increased demand for devices that count cyclists and pedestrians followed because cities can grow strategically based on the collected data. There already exist multiple types of counting devices: camera-based, PIR sensor-based, inductive, or pneumatic. The thesis aims to develop a new approach based on a single-point LiDAR sensor and machine learning algorithms for classification. The system should distinguish between a cyclist, a pedestrian, and other objects. The student will collect data, create a labeled dataset, and compare various classification methods based on deep learning. The most promising classification methods will be incorporated into a counting algorithm and implemented on an STM32 microcontroller.