Real-Time Object Detection Using Convolutional Neural Networks for Autonomous Vehicles

This is where timely and accurate detection by autonomous cars becomes vital. Due to this, Convolution Neural Networks have gained the utmost popularity in object detection and are showing great potential in performing all sorts of sort's r vision tasks. In this project, we propose a Convolution Neural Networks based real-time object detection system for autonomous vehicles. Our system detects and localizes real-time objects based on deep neural network architecture. The function of our software is to take an image from the sensors, like cameras or Liar, on board this car. This image is further input to CNN, which takes these features, extracts the object, and classifies them into a scene. As we previously mentioned, CNN is trained with a large dataset of labeled images, allowing it to recognize hundreds or thousands of different objects accurately. This is then computed further in the CNN and attempts to find bounding boxes around detected objects, giving where that object exists spatially in a scene. Our system's response time has been designed for real-time analysis and should be able to analyze images and detect objects in about milliseconds. This, in return, enables autonomous vehicles to make immediate decisions when considering these objects and navigate safely. Our system allows for real-time object detection, opening the door to high-performance and safe autonomous vehicles that bring us closer than ever before to achieving full autonomy driving.