Relevant Links: Git Repository, Jupyter Notebook
Key Words: Hough Transform, Canny Edge Detection, Region Masking
I created an image processing pipeline to locate lane lines from a car dash camera. This pipeline was then proved on two different image streams.
Relevant Links: Git Repository, Jupyter Notebook
Key Words: Convolutional Neural Network, Data Set Augmentation, Neural Network Model Evaluation, Tensorflow
I created a convolutional neural network to be able to classify different types of german traffic signs. The data set contained 43 different traffic sign classes. The model had a validation accuracy of 95% over 4410 validation images and a test accuracy of 94% over 12630 test images.
Relevant Links: Git Repository, Main Project Script
Key Words: Convolutional Neural Network, Keras, Data Set Augmentation
The project included use of a car driving simulator. The simulator records as you drive the car around circular tracks and pairs images with corresponding steering angles. This was used to create a data set to drive the car autonomously. After training the covnet model, the car was able to drive itself successfully around the track without user input.