This is an overview of self driving car term 1 projects that I completed for the Udacity Self-Driving Car Engineer Nanodegree.
Term 1 covers computer vision using traditional methods as well as with neural networks.
The projects use either OpenCV or TensorFlow. See the individual README of each project for setup details.
See the project writeup ./LaneDetection/LaneDetectionWriteup.md
See also the videos the lane detection here: LaneDetection/test_videos_output/
Use OpenCV to detect lane lines in video from a mounted camera of a driving car.
- See the project writeup ./TrafficSigns/project_writeup.md.
- Built convolutional neural networks to classify traffic signs, and reach 97.55 test accuracy.
- Used image augmentation and batch normalization to improve the model.
- See the project writeup in writeup_report.md
- See the video.mp4 for a test drive based on the trained network (in a simulator).
- Used Keras to build a convolutional neural network that predicts steering angles using video images.
-
See ./LaneAdvanced/advanced_lane_writeup.md
-
See the video ./LaneAdvanced/output_video/project_video_out.mp4
-
Used OpenCV to undistort camera images, used perspective warp to generate an aerial view of the road.
-
Used gradients and color spaces to detect lane lines and also detect the entire lane within lane lanes.
- See the project writeup in ./VechicleDetection/project_writeup.md
- See the videos in ./VehicleDetection/output_video
- Used histogram of gradients to extract features from images.
- Used a rolling window and SVM to classify areas of the image as a car or not.
- Tracked detected images over time in video.