In this recipe, you will learn how to train a custom neural network from scratch using PyTorch and use the model's predictions to classify traffic signs. We shall use the German Traffic Sign Recognition Benchmark (GTSRB) dataset (https://sid.erda.dk/public/archives/daaeac0d7ce1152aea9b61d9f1e19370/published-archive.html) as the input images for training/testing. The images are labeled with 43 different traffic signs. The dataset contains 39,209 training and 12,630 test images. For this recipe, it is also recommended that you use a computer with GPU(s) in it to make the training process faster.
Classifying traffic signs using a deep learning model (with PyTorch)
Getting ready
First, download the compressed pickled dataset...