filtered = is_filtered() start_time = timer() print("Collecting Dataset...") if filtered: # Split the dataset in 0.8 train, 0.1 test, 0.1 validation with shuffle (optionally seed) X_train, X_val, X_test, Y_train, Y_val, Y_test = get_dataset_reshaped(seed=100) else: # Slit the dataset with the same indexes used in the paper (Only for CullPDB6133 not filtered) X_train, X_val, X_test, Y_train, Y_val, Y_test = get_resphaped_dataset_paper() end_time = timer() print("\n\nTime elapsed getting Dataset: " + "{0:.2f}".format((end_time - start_time)) + " s") net = model.CNN_model() #load Weights net.load_weights("NewModelConvConv-best.hdf5") scores = net.evaluate(X_test, Y_test) #print(scores) print("Loss: " + str(scores[0]) + ", Accuracy: " + str(scores[1]) + ", MAE: " + str(scores[2])) CB_x, CB_y = get_cb513() cb_scores = net.evaluate(CB_x, CB_y) print("CB513 -- Loss: " + str(cb_scores[0]) + ", Accuracy: " + str(cb_scores[1]) + ", MAE: " + str(cb_scores[2]))
import numpy as np from keras import optimizers, callbacks from timeit import default_timer as timer from dataset import get_dataset, split_with_shuffle, get_data_labels, split_like_paper, get_cb513 import model dataset = get_dataset() D_train, D_test, D_val = split_with_shuffle(dataset, 100) X_train, Y_train = get_data_labels(D_train) X_test, Y_test = get_data_labels(D_test) X_val, Y_val = get_data_labels(D_val) net = model.CNN_model() #load Weights net.load_weights("Whole_CullPDB-best.hdf5") predictions = net.predict(X_test) print("\n\nQ8 accuracy: " + str(model.Q8_accuracy(Y_test, predictions)) + "\n\n") CB513_X, CB513_Y = get_cb513() predictions = net.predict(CB513_X) print("\n\nQ8 accuracy on CB513: " + str(model.Q8_accuracy(CB513_Y, predictions)) + "\n\n")