/NumericalAnalysisFinal

Primary LanguagePython

Optical Character Recognition

Packages

Python v3.6.1

  1. PyTorch
  2. PIL
  3. NumPy
  4. TorchVision

Train.py

Uncomment line 162 to train model using the train function Train: Parameters

  1. images_path : File path to .npy file containing all image files to be used in training [string]
  2. labels_path : File path to .npy file containing training labels for the provided traing images [string]
  3. validation_percent : validation split percetange to be used during traing [float] initially set to 0.2
  4. epochs : Number of of epochs for training [int] initially set to 30
  5. model_name : File name in which trained model will be saved in. Must in in .pth [string]

Test.py

Uncomment line 26 to test model. Use provided Final_Model.pth file for model_path parameter to test highest performing model trained in a Cloud GPU server Parameters:

  1. path_x : path to images file to test
  2. model_path : path to saved neural network model to be used for testing. Must end in .pth [string]
  3. certainty_threshold : confidence threshold to be used to unknown image detection [int] originally set to 5

Generates a .npy file called predicted.npy containing all predicted labels RETURNS: vector of predicted labels as a NumPy array