Due to a rise in drivers who ignore the speed limits and other traffic regulations, accidents have become more common. To get over the limitations of existing traffic sign detection datasets, which are not universally applicable across nations and locations, we synthesis training data using photos of traffic signs on the road. This technique is used to create a library of synthetic images for recognizing traffic signs in a variety of lighting situations. Using this dataset and an ideal Convolutional Neural Network (CNN), we can create a traffic sign identification and detection system with excellent performance in training and recognition tasks and pinpoint accuracy in detection. This not only reduces the likelihood of accidents but also frees up the driver's attention to focus on the road. The goal of this work is to provide a practical approach to traffic sign detection and identification in India. We presented approaches such neural network and feature extraction to increase the effectiveness in identifying traffic signs and also minimize road accidents, therefore overcoming the limits of current systems. The use of traffic signs is essential for safe driving and the avoidance of accidents, injuries, and deaths. Any ITS will have traffic sign management, which includes automated detection and identification. The need for autonomous cars to automatically recognize traffic signs is paramount in today's advanced technological landscape. In this research, we offer an autonomous system for understanding Indian traffic signs that is based on deep learning. Convolutional Neural Network (CNN)-based end-to-end learning was the inspiration for the automated traffic sign detection and identification system. An original dataset of 6480 photos representing 7056 occurrences of Indian traffic signs organized into 87 categories was used to evaluate the proposed idea. We improve upon the Mask R-CNN model by introducing a number of structural and informational enhancements. We have studied difficult classes of Indian traffic signs that have not been included in any earlier publications.