inference script skeleton

Author: Yasmin Mzayek

inference.py

@@ -8,80 +8,39 @@
 
 
 def main():
-    model_path = "/data/p301081/astronomy/Models"
-
-    filepathTest = "/Trainingdata/20pix_centered_test.csv"
-    print("Loading test data from: ", filepathTest)
-    testSetX, testSetY = cu.load_data_from_csv(filepathTest, shuffle=False, only_positive=False, magRange=[20, 26])
+    
+    image_path = "/data/pg-ds_cit/Projects/Astronomy/AstronomyProject/Images"
+    print("Loading test data from: ", image_path)
+    testSet, X_test, Y_test = cu.load_data_from_images(image_path, 'test')
 
     # Make sure data is float32 to have enough decimals after normalization
     X_test = testSetX.astype('float32')
     # Normalize pixel values between 0 and 1
-    X_train /= 2**8
     X_test /= 2**8
 
-    # If subtract pixel mean is enabled
-    subtract_pixel_mean = False
-    if subtract_pixel_mean:
-        X_train_mean = np.mean(X_train, axis=0)
-        X_train -= X_train_mean
-        X_test -= X_train_mean
-
-    Y_train = trainSetY[:, 0:5]
-    Y_test = testSetY[:, 0:5]
-
-    Y_trainLabels = trainSetY[:, 0]
-    Y_testLabels = testSetY[:, 0]
-
     # input image dimensions
-    img_rows, img_cols = X_train.shape[1:3]
+    img_rows, img_cols = X_test.shape[1:3]
 
     # Convert to correct Keras format
-    X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)
     X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)
 
     print()
-    print('Data loaded: train:', len(X_train), 'test:', len(X_test))
-    print('trainSetX:', trainSetX.shape)
-    print('X_train:', X_train.shape)
-    print('trainSetY:', trainSetY.shape)
-    print('Y_train:', Y_train.shape)
+    print('Data loaded: test:', len(X_test))
+    print('X_test:', X_test.shape)
 
-    if resnet:
-        model = load_model(modelName, custom_objects={'custom_YOLO_loss': cu.custom_YOLO_loss,
-                                                      'f1_metric': cu.f1_metric})
-    else:
-        model = load_model(modelName, custom_objects={'custom_loss': cu.custom_YOLO_loss})
+    model = load_model(model_path)
 
     print(model.summary())
 
-    # Evaluate
-    scoresTrain = model.evaluate(X_train, [Y_train, Y_trainLabels], verbose=2)
-    scoresTest = model.evaluate(X_test, [Y_test, Y_testLabels], verbose=2)
-    print(scoresTrain)
-    print(scoresTest)
-
     # Make predictions
-    predictionsTrain = model.predict(X_train)
     predictionsTest = model.predict(X_test)
-
-    # ResNet predictions array is different shape that simple CNN
-    if resnet:
-        predictionsTrain = predictionsTrain[0]
-        predictionsTest = predictionsTest[0]
-
-    print("\nTraining set:")
-    cu.analyze_5unit_errors(predictionsTrain, Y_train)
-    print("\nTest set:")
-    cu.analyze_5unit_errors(predictionsTest, Y_test)
-
-    dataframe2Dtrain = cu.create_histogram_2d(predictionsTrain, trainSetY, binsMag, binsLength)
-    cu.plot_results_heatmap(dataframe2Dtrain, binsMag, fig_name = 'Plot_2D_histogram_CNN_train.pdf', title="Deep learning training set completeness") #ym
-
-    dataframe2Dtest = cu.create_histogram_2d(predictionsTest, testSetY, binsMag, binsLength)
-    cu.plot_results_heatmap(dataframe2Dtest, binsMag, fig_name = 'Plot_2D_histogram_CNN_test.pdf', title="Deep learning test set completeness") #ym
-
-    plt.show()
+    predictionsTest = [round(pred[0]) for pred in predictionsTest]
+    
+    # Evaluate
+    if Y_test:
+        scoresTest = model.evaluate(X_test, Y_test, verbose=2)
+        print(scoresTest)
+        test_set_metrics = cu.analyze_5unit_errors(predictionsTest, Y_test)
 
 
 if __name__ == "__main__":