License Plate Recognition (LPR) has a wide variety of uses and a very useful technological skill in today's world. The license plate is vehicle's singular identifier, and it is used to gather information about the owner of the vehicle to lessen traffic law violations, pay fines for infractions, addition to the security uses. Therefore, the aim of study is license plates must be detected and identified with highest accuracy and the least loss. The major processes of the proposed system are license plate detection and character recognition. Using various base models and configurations, in proposed system we used Detectron2's Faster R-CNN network with two models ResNet (R101-FPN) and ResNeXt(X101-FPN) for feature extractor to plate detection. After preprocessing and some enhance operations, Teseract OCR system would be used as character recognition module. The system results achieved Based on the detection performance comparison between both models, R101-FPN accuracy was 99.2%, while the accuracy for X101-FPN was 99.6%. Although there are many challenges as distorted, dark, and high-light images, X101-FPN gave better prediction results than R101-FPN, although its slower. The experimental findings demonstrate the efficiency of the proposed NPRs framework that includes the (X101-FPN) base model for Detectron2 Faster R-CNN and pytesseract OCR has high accuracy up to 99.6%.