Abstarct: Management, control and monitoring of vehicle traffic is one of the important issues that if it is not done correctly, it will have traffic, security and social consequences. One of the most important aspects of car traffic monitoring is the identification of their license plates, so that it can be done with the least time and the most accuracy, considering the speed and volume of traffic. Before the emergence of deep neural networks, traditional machine vision techniques were used to process images and identify and register vehicle license plates, which often had many errors. With the emergence of deep networks and the evolution of object recognition in the image, traditional techniques gave way to deep networks, and networks such as Yolo were able to cope well with the challenges of license plate recognition. But the problem that has limited the commercial use of such algorithms is the need for high processing resources, which has caused the creation of expensive license plate recognition systems. In this research, we intend to solve this problem by relying on the Yolo object detection algorithm, by proposing an agile and high-precision network, with the aim of reducing processing time and minimal use of processing resources, in such a way that license plate recognition systems don’t have require expensive graphics processors. The result of this research is the introduction of a convolutional network with 24 laxyers and two outputs, which has been able to provide similar and even higher accuracy than the tiny versions of the Yolo network with a significant reduction in processing time and volume. About 19,000 car images and 40,000 cropped license plate images were used to train the network, and among the images of both stages, there were images with distorted license plates and images rotated up to 40 degrees. The accuracy of the above network to recognize the license plate in the image is 97% with the recall criterion, and 98% to recognize the characters from the cropped image of the license plate. Also, in terms of operation volume, the proposed network has 7.5 and 9 times less operation volume than the smallest version of Yolo 4 and Yolo 7 networks, respectively, and it is three to four times faster in terms of processing time.