In order to autonomously control agricultural vehicles, one potential solution could be to follow driving lanes left by tractors, as these are almost always present and permanent inside fields. This is because the simple use of limiting driving to permanent lanes increases yield by an average of $5-15\%$. Different sensors can be used to follow these driving lanes, in this project however, a traditional RGB camera was used. A comparison of the performance of a classic computer vision system and a neural network's ability to find driving lanes on agricultural fields, is given. Although this given use-case is fairly niche, the same logic could apply to other areas of expertise. A complete classic computer vision system was developed from scratch, while the neural network is an implementation of YOLOv3. Both of the approaches yield decent results and they are both able to find driving lanes. The classic computer vision system almost always provides a result which could be used to guide an agricultural vehicle. However, this system is more prone to faulty outputs than the neural network. The neural network often does not predict anything inside the inference images, although, it almost never makes false predictions. From this comparison, it was concluded that a classic computer vision system is more appropriate to use given a small time frame or if the implementation serves as a proof of concept. The neural network approach, given more training or a larger dataset, would be better suited for a large scale project because of the more accurate results.