Behavioral Cloning

Project idea

Steering angle

Overview

Data

• Collecting the data
• Data includes large amount of images
• Data from real human driving

Model

• Design, train and validate a model that predicts a steering angle from image data
• Using neural network
• What is neural network

Layer in Neural Network

• Convolutional Layer
• Fully Connected Layer(Dense)
• Flatten

Fully connected layer

• This layer takes an input and output an N dimensional vector.
• Based on the activation map like curve detector with high value,
• The layer take the previous layer and decides which features correlate to a specific class, in self-driving, it correlates to the different angle. Similiarly, if the program is predicting that angle is 0.5, it will have high values in the activation maps that represent high level features.

Flatten

• Transforming multiple-dimensional into a 1D feature vector
• Flatten layer is between Conv and fully connected layer.
• Fully connected layer only take a large piece of unstructured data.

Training

• The way the computer is able to adjust its filter values (or weights) is through a training process called backpropagation.
• Backpropagation can be separated into 4 distinct sections, the forward pass, the loss function, the backward pass, and the weight update
• This idea of being given an image and a label(angle) is the training process that CNNs go through.

validate

• TRAINING SET: It is used to adjust the weights on the neural network.
• VALIDATION SET/Test phase: in order to estimate how well your model has been trained (that is dependent upon the size of your data, the value you would like to predict, input etc) and to estimate model properties (mean error for numeric predictors, classification errors for classifiers, recall and precision for IR-models etc.) ,minimize overfitting

Test

• After getting real model, test it in the simulator