Classification results of ML models with different inputs across five traffic sign datasets

A dataset of the classification results for five diverse ML models with five diverse inputs. The classification task is based on five existing traffic sign datasets. This dataset consists of five sheets, each sheet consists of 26 columns representing the ground truth class label and classification result labels. ML models are LeNet, AlexNet, ResNet50, VGG16, and ResNet18. Different inputs are original data, noise-added data (variance is 0.01^2), rotated data (5 degrees counterclockwise), shifted down (by 1 pixel) data, and shifted right (by 1 pixel) data. The five traffic sign datasets are the Chinese Traffic Sign Dataset (CTSD), the German Traffic Sign Recognition Benchmark (GTSRB), Traffic Sign Classification Dataset (TSCD), Turkey Traffic Sign (TTS) and Arabic Traffic Signs (ATS). The numbers of samples are 1994, 12630, 14628, 5313, and 9240, for CTSD, GTSRB, TSCD, TTS, and ATS, respectively.