Prevention Detection for Cyclists based on Faster R-CNN
Ryan Teng · 2022
Cycling is becoming increasingly common on the roads, and the growth of this activity raises safety concerns. Bicycles are becoming one of the most common ways of transportation worldwide. Almost every day, though, there is an incident in which a driver hits and kills a rider. In the past few years, the number of avoidable deaths from bicycle transportation events has climbed by 37%. (National Safety Council). Thus, prediction prevention is an essential task in order to avoid such incidents. This article provides a summary of recent developments in commercial devices for cyclists as well as intent estimation to improve rider safety. Understanding the intentions of vehicles enables cyclists to take precautionary measures to prevent collisions. To accomplish this, the development of methods/ techniques for the autonomous vehicle, such as machine learning (ML), will be investigated. For example, employing DL techniques like R-CNN, and Faster R-CNN, the development of pedestrian detection has improved dramatically. While machine learning techniques have been around for decades, the technology required to implement them has just recently been made available. However, using a different approach by employing these ML approaches for vehicle detection and applying it for prediction estimations can allow for an effective and accurate way of vehicle intent estimation. A standard generic object identification framework is used to recognize the vehicle and assess the speed and distance from the cyclist. Then warn the rider ahead of time if it is speeding up or slowing down, calculating if it can slow down at that pace. To tackle the given challenge, we employed a two-stage Faster R-CNN detector.