Model

Step 1:Collect data

1.Program the code to control car.

2.Import the control code and get images when car is running on the track.

Step 2:Transfer the dataset from JPG to NPZ

1.Compress the images from 3594 to 14.

2.Returns the data matrix of the images and the corresponding tag value.

Step 3: Build and train model

1.Load data and split into training and validation sets. Match all eligible files and return them as a list.

2.Build the End-to-End model.

3.Train the model with 'min' mode (if the detected value stops falling, the training is aborted), Save the best model.

Step 4: Load the model and achieve auto-driving

1.Acquire new images (dataset) by Raspberry Pi camera.

2.Predict the maximum possibility of directions by using neural network models.

3.Control the car according to possibilities.

Google Sites

Report abuse