Upload app.py

app.py

@@ -1,68 +1,58 @@
-import numpy as np
-import gradio as gr
-import tensorflow as tf
-import matplotlib.pyplot as plt
-from matplotlib import cm
-from PIL import Image
-import pandas as pd
-
-def normalise_data(data, min_val, max_val, low=0, high=1):
-    '''
-    Normalises the data to the range [low, high]
-
-    Parameters
-    ----------
-    data: numpy array, data to normalise
-    low: float, minimum value of the range
-    high: float, maximum value of the range
-
-    Returns
-    -------
-    normalised_data: float, normalised data
-    '''
-    normalised_data = (data - min_val)/(max_val - min_val)
-    normalised_data = (high - low)*normalised_data + low
-    return normalised_data
-    
-x_test_1 = np.load('./data/x_test_1_LFP.npy')
-x_test_1 = normalise_data(x_test_1, 0, 1, 0, 100)
-x_test_1 = x_test_1.reshape(-1, 6, x_test_1.shape[1])
-
-x_test_DTW_1 = np.load('./data/x_test_DTW_1_LFP.npy')
-x_test_DTW_1 = x_test_DTW_1.reshape(-1, 6, x_test_DTW_1.shape[1], x_test_DTW_1.shape[2], x_test_DTW_1.shape[3])
-
-model = tf.keras.models.load_model('./models/model-bestLFP_V2.h5',compile = False)
-
-def predict(Cell_number, Duty_Cycle, Cycle_number):
-  # ------------------------ Prediction ------------------------
-  # select cell data
-  data = x_test_1 #if Cell_number == '1' else x_test_2 if Cell_number == '2' else x_test_3
-  data_DTW = x_test_DTW_1 #if Cell_number == '1' else x_test_DTW_2 if Cell_number == '2' else x_test_DTW_3
-  # select cycle number
-  cycle = 0 if Cycle_number == '10' else 1 if Cycle_number == '50' else 2 if Cycle_number == '100' else 3 if Cycle_number == '200'else 4 if Cycle_number == '400' else 5
-
-  sample = data[Duty_Cycle-1][cycle]
-  sample_DTW = data_DTW[Duty_Cycle-1][cycle]
-  prediction = model.predict(np.expand_dims(sample_DTW, axis=0))
-  pred = {"LLI ": str(prediction[0][0]), "LAMPE ": str(prediction[0][1]), "LAMNE ": str(prediction[0][2])}
-  # --------------------------- ICA + image----------------------------
-  
-  ICA_reference = x_test_1[0][0]
-  d = {' ': np.linspace(1, len(ICA_reference), len(ICA_reference)), 'pristine': ICA_reference, 'degraded': sample}
-  df = pd.DataFrame(data=d)
-
-  image_array=sample_DTW.reshape(sample_DTW.shape[0], sample_DTW.shape[1])
-  image_array = normalise_data(image_array, np.min(image_array), np.max(image_array))
-  im = Image.fromarray(np.uint8(cm.inferno(image_array)*255))
-  
-  return pred, df, im
-  
-iface = gr.Interface(
-	fn=predict,
-	inputs=[gr.inputs.Radio(["Cell #1", "Cell #2", "Cell #3"]), gr.inputs.Slider(1, 1000, step=1), gr.inputs.Radio(["10", "50", "100", "200", "400", "1000"]), "checkbox"],
-	title="LFP degradation diagnosis",
-	description="Enter cell number, duty cycle and cycle number to predict the percentage of LLI, LAMPE and LAMNE",
-	outputs=[gr.outputs.Label(label="Prediction"), gr.outputs.Timeseries(x=" ", y=["pristine", "degraded"]), gr.outputs.Image(type='pil', label="DTW image")],
-	allow_screenshot=False,
-	layout="unaligned")
+import numpy as np
+import gradio as gr
+import tensorflow as tf
+import matplotlib.pyplot as plt
+from matplotlib import cm
+from PIL import Image
+import pandas as pd
+from dtaidistance import dtw
+
+def getDTWImage(IC_reference, sample, size):
+	d, paths = dtw.warping_paths(IC_reference, sample, window=int(size/2), psi=2)
+	x = np.array(paths)
+	# mask values that are not filled
+	x = np.where(x == np.inf, -99, x)
+	# negative values are replaced by 0
+	x = np.where(x < 0, 0, x)
+	# normalise values
+	x = x/np.max(x)
+	# reshape the array
+	x = np.expand_dims(x, -1).astype("float32")
+	return x
+
+data = np.load('./data_LFP.npy')
+model = tf.keras.models.load_model('./models/model-bestLFP_V2.h5',compile = False)
+
+def predict(Cell_number, Duty_Cycle, Cycle_number):
+  # ------------------------ Prediction ------------------------
+  # select cell data
+  # data = x_test_1 #if Cell_number == '1' else x_test_2 if Cell_number == '2' else x_test_3
+  # data_DTW = x_test_DTW_1 #if Cell_number == '1' else x_test_DTW_2 if Cell_number == '2' else x_test_DTW_3
+  # select cycle number
+  cycle = 0 if Cycle_number == '10' else 1 if Cycle_number == '50' else 2 if Cycle_number == '100' else 3 if Cycle_number == '200'else 4 if Cycle_number == '400' else 5
+
+  IC_reference = data[0][0]
+  sample = data[Duty_Cycle-1][cycle]
+  sample_DTW = getDTWImage(IC_reference, sample, size)
+  prediction = model.predict(np.expand_dims(sample_DTW, axis=0))
+  pred = {"LLI ": str(prediction[0][0]), "LAMPE ": str(prediction[0][1]), "LAMNE ": str(prediction[0][2])}
+  
+  # --------------------------- ICA + image----------------------------
+  d = {' ': np.linspace(1, len(ICA_reference), len(ICA_reference)), 'pristine': ICA_reference, 'degraded': sample}
+  df = pd.DataFrame(data=d)
+
+  image_array=sample_DTW.reshape(sample_DTW.shape[0], sample_DTW.shape[1])
+  image_array = normalise_data(image_array, np.min(image_array), np.max(image_array))
+  im = Image.fromarray(np.uint8(cm.inferno(image_array)*255))
+  
+  return pred, df, im
+  
+iface = gr.Interface(
+	fn=predict,
+	inputs=[gr.inputs.Radio(["Cell #1", "Cell #2", "Cell #3"]), gr.inputs.Slider(1, 1000, step=1), gr.inputs.Radio(["10", "50", "100", "200", "400", "1000"]), "checkbox"],
+	title="LFP degradation diagnosis",
+	description="Enter cell number, duty cycle and cycle number to predict the percentage of LLI, LAMPE and LAMNE",
+	outputs=[gr.outputs.Label(label="Prediction"), gr.outputs.Timeseries(x=" ", y=["pristine", "degraded"]), gr.outputs.Image(type='pil', label="DTW image")],
+	allow_screenshot=False,
+	layout="unaligned")
 iface.launch(share=True)