=================================== README Eeshaan Sharma - 2015CSB1011 Shivam Mittal - 2015CSB1032 =================================== NOTE - Q1), Q2) and the competition part, all are implemented in MATLAB Q1) Warm up Exercise How To Run : 1.) Navigate to the following directory - /code/Q1 2.) In the Matlab window type - l31 3.) Follow the input prompt to run on various values of number of epochs, learning rate, number of neurons in hidden layer. Q2.1) Predicting the Steering Angle NOTE - Q2.1) requires the steering folder to be present in code folder How To Run : 1.) Navigate to the following directory - /code/Q2_1 2.) In the Matlab window type - Q2_1 3.) Follow the input prompt to run on various values of number of epochs, learning rate, mini batch size and dropout probability of different layers. NOTE - The images present in steering folder will be read on first run of the code and then the input matrix X (1 X 1024) will be stored in Images.csv file. Q2.2) Competition part The output for the test images is given in the text file test_output.txt Note - For this part, we have extracted features from the images, both for the training/validation images and the test images. Also, our model is an ensemble of 3 neural networks, so there are 3 weight files corresponding to each neural network. Download all the files from the google drive folder : https://drive.google.com/drive/u/0/folders/1mv5iiffdhE0Mdfi4fEhfyznOzR8CxvfL and include these in the folder Q2_2. Otherwise you'll have to wait a very very long time for 3 neural nets to train. If you do not have the given files, and want to train the network again then, steering folder and l3-test folder should be in the parent directory (code) of Q2_2. Also note, if you want to test on some different images, then delete the features_test.csv file and put those images and text file in l3-test folder. Running/compiling ------------------ 1.) Navigate to the following directory - /code/Q2_2 2.) In the Matlab window type - main
picnkname/Predicting_steering_angle
Neural network to predict the steering angle the road image for a self-driving car application that is inspired by Udacity’s Behavior Cloning Project
MATLAB