How to use our code

To train the CNN:

• To use our dataset to train and get weight
  input: python3 train.py --save 1
• To train and display validation accuracy
  input: python3 train.py --save 0

To modify the CNN parameters:

• Add your own dataset
  (default = the dataset we provided):
  --data_path your_dataset_directory_path
• Change the batch size
  (default = 4):
  --batch_size your_batch_size
• Change the learning rate
  (between 0.0 to 1.0, default = 0.001):
  --learning_rate your_learning_rate
• Change the number of epoches
  (default = 200):
  --epoches your_epoches

Finally, you can test on your Cozmo:

• get TurnToCube.py with command: genfsm TurnToCube.fsm
• In simple_cli, input runfsm('TurnToCube')

Source Code

• CubeClassifiierNet.py
  
• train.py
  
• TurnToCube.fsm
  
• weight