Udacity Students Review the Self-Driving Car Program
Joshua provides a concise, intuitive explanation of Kalman filters and extended Kalman filters, with just enough math to make it seem real:
“In our case of tracking the position and velocity of an object, we could use laser or radar range sensors. However, while laser sensors represent measurements in cartesian coordinate, the radar sensors represent in polar coordinate. A direct conversion of polar coordinate to cartesian coordinate gives nonlinearity and so Kalman filter is no more useful. Hence, we have to be able to linearly map from the polar coordinate to the cartesian coordinate. EKF uses the method called first order Taylor expansion to obtain linear approximation of the polar coordinate measurements in the update. In this process, a Jacobian matrix is produced, which represents the linear mapping from polar to cartesian coordinate, applied at the update step.”
Martijn has some fun images from his advanced lane-finding image, including an error image that looks a little bit like modern art:
This code created the amazing plot showed below by accident, might be fun :)
image = mpimg.imread(‘test_images/test6.jpg’) thresh = (180, 255) gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY) binary = np.zeros_like(gray) binary[(gray > thresh) & (gray <= thresh)] = 1 plt.plot(binary) plt.show()
Mehdi provides a concise summary of his vehicle detection pipeline, as well as some thoughts about other approaches for solving the problem:
“The pipeline is likely to fail if :
Worse weather/lighting conditions.
Less contrast between the cars and the road in general.
Going up/down hills.
We used classical computer vision techniques and added a SVC classifier in this project. It’s really interesting to understand and get familiar with these concepts.
But again, I’d be very interested to see how deep learning would tackle this problem.”
Mithi has a nice description of many different parts of the Nanodegree Program — the projects, the instructional material, the career development modules — and is very complimentary. We’re blushing :-)
“One of the most simple yet the most exceptional thing for me is that the platform also asks you to rate and give your feedback on almost any small interaction, from the videos to the project reviews.”
Ashis provides a terrific summary of each project in the program, what he learned, and how much he liked the process:
“The Behavioral Cloning Project is the second and last project in deep learning and the objective is to drive the car on the track safely (prediction of steering angle). I really enjoyed my time spent during this project . It was a great theoretical and practical learning experience in many ways.”