Enforce Fines For Red Light Violations And White Line Crossing

---

Abstract

According to recent assessments, traffic offenses have mostly resulted in an increase in fatalities and injuries on Indian roadways. Because it is laborious to manually identify the automobiles that are in violation of the traffic laws, an automated computer vision-based object detection model was required for this task. This paper's main idea is to use a single video frame to identify several infractions. To perform various operations, the input video stream from the security camera is analyzed and annotated. The COCO dataset is utilized for red-light leaping, while the annotation of Google image photos creates the over- boarding dataset. Tensorboard is used to visualize the results after the model has been trained. F-measure, P-measure, Precision, and Recall are the parameters that are used. Red light skipping has a 93% accuracy rate, while overboarding has a 0.5:0.95 mAP value. In order to identify different violations, this system makes the most of the video stream.

---

Keywords

Convolutional Neutral Networks (CNN) Support Vector Machine (SVM), Mobile Net SSD Algorithm, Human Activity Recognition