/AI_Car_and_Pedestrian_Detector_System

An AI based system that do detects the cars and the pedestrians from a captured video.

Primary LanguagePythonMIT LicenseMIT

AI Car and Pedestrian Detector System

An AI based system that do detects the cars and the pedestrians from a captured video.

Requirements

Apart from installing Python, you'll also need to pip and install the OpenCV, and this is pretty much it. Optionally you'll need to install the numpy because some of the OpenCV methods return numpy arrays so that's why.

The Trainig Data

The trainging datasets are downloaded from the internet you can find the car's dataset also known as car_detector data set, and the pedestrian's dataset known as haarcascade_fullbody.

The Code Flow

  • First and foremost import the opencv liberary that's gonna do a lot of work for us in this detection system.
  • Save the captured video file in variable video
  • Save the car and pedestrian pre-trained classifiers and save it in car_tracker_file and pedestrian_tracker_file respectively
  • Create the classifiers for the both, by the use of opencv cascade classifiers
  • Now utill the video, (our simulated car) didn't stops, the process will continue
    • And the process is, to read the video frames
    • If it reads successfully convert that frame to the gray-scale or say black and white, because it can be process faster than in the colored form
    • Now detect the cars and pedestrian from the frames, which it gives us the an array of 4D arrays, of each and every single car and pedestrian.
    • Now for every car and pedestrian draw a rectangle over it
    • Show the frames by cv2.imshow() method
    • And at the last release the VideoCapture object we've created