Using computer vision and machine learning to classify traffic signs is a sophisticated endeavor to pioneer new Advanced Driver Assistance Systems (ADAS) and enhance existing ones. For this purpose, this article conducts an experimental study using the following steps: i) Establish a dataset comprising Ecuadorian regulatory traffic signs. ii) Four classification algorithms are employed based on the well-known CNN architectures. iii) Train four different CNN models and select the most optimal one among them. Experimentally, the CNN1 model has been selected as the optimal choice, showing superior performance and producing remarkable results in standard classification metrics, such as a \(98.6\%\) accuracy, a \(99.30\%\) sensitivity, a \(99.63\%\) AUC, a \(98.89\%\) AP, and a \(99.42\%\) F1 score. Ultimately, this proposal shows impressive efficiency, processing images sized at \(32 \times 32\) pixels on a Jetson Nano mobile device in just 180 milliseconds.