if len(sys.argv) <= 2: sys.exit("Specify the patient to process, e.g. 'LIDC-IDRI-0001'.") patId = sys.argv[1] imageType = sys.argv[2] # bin or orig # If pickled dataset joining X & y values doesn't exist, then create it; else, use it. pickle_file_name = "Evans-MacBook-Pro.local-8x8_edge_patches-"+imageType+"-"+patId+".pickle" if os.path.isfile(pickle_file_name): input_data_LIDC.esprint("Unpickling: " + pickle_file_name) with open(pickle_file_name, "rb") as pickle_file: dataset_input = pickle.load(pickle_file) else: dataset_input = input_data_LIDC.read_data_sets( '../../../LIDC_Complete_20141106/LIDC-IDRI-edge_patches/'+patId, (os.sep + imageType + os.sep), '*.tiff', '../../../LIDC_Complete_20141106/Extracts/master_join4.csv', '../../../LIDC_Complete_20141106/Extracts/DICOM_metadata_extracts/', '*.csv') input_data_LIDC.esprint("Pickling: " + pickle_file_name) with open(pickle_file_name, "wb") as pickle_file: pickle.dump(dataset_input, pickle_file) esprint("Done pickling.") # Randomize the image & label set in-place, taking care to maintain array correspondance. # First, re-merge the training, validation, and test sets into a single set. train_images, train_labels = dataset_input[0] # validation_images, validation_labels = dataset_input[1] test_images, test_labels = dataset_input[1]
# y_len = 5 ### Read LIDC images and labels ### # Memoization: If we previously joined images to malignancy values and saved the results to a pickle file, then load that pickle file. pickle_file_name = cfg["pickle_file_name"] if os.path.isfile(pickle_file_name): input_data_LIDC.esprint("Unpickling: " + pickle_file_name) with open(pickle_file_name, "rb") as pickle_file: dataset_input = pickle.load(pickle_file) # Else, we haven't computed join, so let's compute it and create the pickle file. This will take a while and use all your CPU cores. else: dataset_input = input_data_LIDC.read_data_sets( cfg["images_dir_path"], cfg["images_file_glob"], cfg["master_join4_path"], cfg["DICOM_metadata_extracts_dir_path"], cfg["DICOM_metadata_extracts_file_glob"]) input_data_LIDC.esprint("Pickling: " + pickle_file_name) with open(pickle_file_name, "wb") as pickle_file: pickle.dump(dataset_input, pickle_file) # Specify the input characteristics for samples (X) and labels (y). img_px_len_x = cfg["image_len_x"] img_px_len_y = img_px_len_x X_len = img_px_len_x * img_px_len_y y_len = cfg["y_len"] # Randomize the image & label set in-place, taking care to maintain array correspondance. # First, re-merge the training, validation, and test sets into a single set.
if len(sys.argv) <= 2: sys.exit("Specify the patient to process, e.g. 'LIDC-IDRI-0001'.") patId = sys.argv[1] imageType = sys.argv[2] # bin or orig # If pickled dataset joining X & y values doesn't exist, then create it; else, use it. pickle_file_name = "Evans-MacBook-Pro.local-8x8_edge_patches-" + imageType + "-" + patId + ".pickle" if os.path.isfile(pickle_file_name): input_data_LIDC.esprint("Unpickling: " + pickle_file_name) with open(pickle_file_name, "rb") as pickle_file: dataset_input = pickle.load(pickle_file) else: dataset_input = input_data_LIDC.read_data_sets( "../../../LIDC_Complete_20141106/LIDC-IDRI-edge_patches/" + patId, (os.sep + imageType + os.sep), "*.tiff", "../../../LIDC_Complete_20141106/Extracts/master_join4.csv", "../../../LIDC_Complete_20141106/Extracts/DICOM_metadata_extracts/", "*.csv", ) input_data_LIDC.esprint("Pickling: " + pickle_file_name) with open(pickle_file_name, "wb") as pickle_file: pickle.dump(dataset_input, pickle_file) esprint("Done pickling.") # Randomize the image & label set in-place, taking care to maintain array correspondance. # First, re-merge the training, validation, and test sets into a single set. train_images, train_labels = dataset_input[0] # validation_images, validation_labels = dataset_input[1] test_images, test_labels = dataset_input[1]
# img_px_len_x = 236 # img_px_len_y = img_px_len_x # X_len = img_px_len_x * img_px_len_y # y_len = 5 ### Read LIDC images and labels ### # Memoization: If we previously joined images to malignancy values and saved the results to a pickle file, then load that pickle file. pickle_file_name = cfg["pickle_file_name"] if os.path.isfile(pickle_file_name): input_data_LIDC.esprint("Unpickling: " + pickle_file_name) with open(pickle_file_name, "rb") as pickle_file: dataset_input = pickle.load(pickle_file) # Else, we haven't computed join, so let's compute it and create the pickle file. This will take a while and use all your CPU cores. else: dataset_input = input_data_LIDC.read_data_sets( cfg["images_dir_path"], cfg["images_file_glob"], cfg["master_join4_path"], cfg["DICOM_metadata_extracts_dir_path"], cfg["DICOM_metadata_extracts_file_glob"]) input_data_LIDC.esprint("Pickling: " + pickle_file_name) with open(pickle_file_name, "wb") as pickle_file: pickle.dump(dataset_input, pickle_file) # Specify the input characteristics for samples (X) and labels (y). img_px_len_x = cfg["image_len_x"] img_px_len_y = img_px_len_x X_len = img_px_len_x * img_px_len_y y_len = cfg["y_len"] # Randomize the image & label set in-place, taking care to maintain array correspondance. # First, re-merge the training, validation, and test sets into a single set. train_images, train_labels = dataset_input[0] validation_images, validation_labels = dataset_input[1] test_images, test_labels = dataset_input[2]