Traffic Sign recognition is a crucial part of current systems of transportation and Advanced Driving Assistance Systems (ADAS). The safe operation of an autonomous vehicle and the effective management of road infrastructure depend on a traffic sign recognition system. Existing techniques to traffic sign identification often face difficulties to classify the traffic signs accurately and resulting in lower reliability in identifying critical traffic signs. This limitation compromises road safety and prevents ADAS from reaching its maximum potential. The research focuses to classify the Indian traffic signs using transfer learning with a pre-trained model. The pre-trained convolutional neural networks (CNNs) such as ResNet50, Xception, DenseNet121, and EfficientNetB5 are used to develop the convolutional and pooling layers that are used in the proposed framework. In order to help with accurate classification of traffic indicators, fully-connected layers as well as an output layer have also been developed. The proposed method is thoroughly tested using the Kaggle Indian Traffic Signal Prediction dataset, which contains 85 distinct types of traffic signals regularly noticed on Indian highways. The test outcomes show that ResNet50 outperformed among the four pre-trained CNNs, with an impressive testing accuracy of 86.47% and a training accuracy of 89.91%.