Inmodern society, road transportation is an essential statutory organ; nevertheless, because of a growth in road accidents, it is responsible for the loss of over a million lives and billions of dollars each year in the worldwide economy. The goal is to offer early warning of prospective accidents so that necessary actions may be taken to limit the frequency and severity of incidents. The system may use multiple data sources to identify and anticipate accidents, such as vehicle speed, braking patterns, road conditions, weather information, and real-time traffic data. If the system detects a probable accident, it may immediately notify emergency services, traffic management authorities, or other relevant parties, allowing them to react more effectively.In the work that has been suggested, detection is carried out using picture frames, and prediction is carried out using metadata. It does this by using a handful of the most recent instance segmentation approaches, such as MaskRCNN for accident detection and Resnet for extracting features from pictures and maps feature to build features for accident prediction models in addition to metadata.In this research, the authors suggested using MaskRCNN for detecting and predicting road accidents by combining Xgbsoot with Resnet101. Both our prediction and detection accuracy comes in at 97.67%. Our prediction accuracy is 93.87%.