Update app.py

app.py

@@ -6,7 +6,6 @@ import numpy as np
 import matplotlib.pyplot as plt
 import gradio as gr
 
-# Creating a numpy array of shape (8, 16, 1)
 flow_field = np.ones((128,256), dtype = np.uint8)
 
 # Changing the left input side
@@ -18,6 +17,14 @@ flow_field[0,:] = 2
 # Changing the bottom layer
 flow_field[-1,:] = 2
 
+mean_u = 0.075003795
+mean_v = -0.000036
+mean_p = 0.004301
+
+std_dev_u = 0.04605
+std_dev_v = 0.013812
+std_dev_p = 0.007917
+
 def nvs_loss(y_pred, rho=10, nu=0.0001): #arbitary rho and nu(Later use values of air)
       u,v,p = tf.split(y_pred, 3, axis=3)
 
@@ -172,7 +179,6 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
     test_img = patch_stiching(flooded_image, h, w, x0, y0)
 
 # Loading and Compiling the Model
-    #model_path = "/content/drive/MyDrive/Pinns_Loss_file.h5"
     model_path = "Pinns_Loss_file.h5"
     model = load_model(model_path, compile = False)
     model.compile(loss=custom_loss, optimizer=tf.keras.optimizers.AdamW(learning_rate = 0.0001), metrics=['mae', 'cosine_proximity'])
@@ -181,6 +187,11 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
     prediction = model.predict(test_img) # (prediction.shape = (1, 128, 256, 3))
     u_pred, v_pred, p_pred = np.split(prediction, 3, axis=3) # shape of u_pred, v_pred, p_pred = (1, 128, 256, 1)
 
+ # De-Normalizing teh Data:
+    u_pred = ((u_pred*std_dev_u) + mean_u)
+    v_pred = ((v_pred*std_dev_v) + mean_v)
+    p_pred = ((p_pred*std_dev_p) + mean_p)
+
  # Making test_img in shape required by zero_pixel_location
     req_img = squeeze_function(test_img)
 
@@ -192,7 +203,10 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
     u_profile = u_pred[0][:,:,0]  # shape of u profile to compatible shape (H, W) = (128, 256)
     v_profile = v_pred[0][:,:,0]
     p_profile = p_pred[0][:,:,0]
-    p_profile[p_profile>1.6] = 1.6
+    p_profile[p_profile>0.02] = 0.02
+    hist, bins = np.histogram(p_profile, bins=20)
+    print(hist)
+    print(bins)
 
 # Creating a copy of the above profiles-
     u_profile_dash = np.copy(u_profile)
@@ -217,7 +231,7 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
     velocity = np.sqrt(u_profile_dash_1**2 + v_profile_dash_1**2)
     ax = plt.subplot()
     ax.imshow(velocity, cmap = 'gray', extent = (0,256, 0,128))
-    q = ax.quiver(X[5::7,5::7], Y[5::7,5::7], u_profile_dash[5::7,5::7], v_profile_dash[5::7,5::7], pivot = 'middle', color = 'red')
+    q = ax.quiver(X[5::5,5::5], Y[5::5,5::5], u_profile_dash[5::5,5::5], v_profile_dash[5::5,5::5], pivot = 'middle', color = 'red')
     ax.quiverkey(q, X=0.9, Y=1.07, U=2,
                 label='m/s', labelpos='E')
     plt.title("Velocity distribution", fontsize = 11)
@@ -226,7 +240,7 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
 
  # StreamLine Plot
     streamline_plot = plt.figure(figsize = (14,6), edgecolor = "gray")
-    plt.streamplot(X, Y, u_profile_dash, v_profile_dash, density = 3.5)
+    plt.streamplot(X, Y, u_profile_dash, v_profile_dash, density = 4)
     plt.axis('scaled')
     plt.title("Streamline Plot", fontsize = 11)
     plt.xlabel("Length of Channel", fontsize = 11)
@@ -248,9 +262,10 @@ def fill_shape_with_pixels(img): #img is taken by gradio as uint8
 with gr.Blocks(theme="Taithrah/Minimal") as demo:
   gr.Markdown(
     """
-    # Physics Constrained DNN for Predicting Mean Turbulent Flows
+    # Channel Flow - Physics Constrained DNN for Predicting Mean Turbulent Flows
     The App solves 2-D incompressible steady state NS equations for any given 2-D closed geometry. Geometry needs to be drawn around the center of the patch.\n
     It predicts the streamlines,horizontal & vertical velocity profiles and the pressure profiles using a hybrid loss function.\n
+    Model Parameters (In SI Units) - Kinematic Viscosity = 0.0001, Input horizontal velocity = 0.075, Input vertical velocity = 0
     """)
   with gr.Row():
     with gr.Column():