added raw activations

app.py

@@ -82,6 +82,37 @@ def get_activations(intermediate_model, image: list,
     
     return output, in_image, activation_1, activation_2
 
+def plot_activations(activation_1, activation_2, origin='lower'):
+    
+    
+     ##### Make the activation figure ######
+    plt.rcParams['xtick.labelsize'] = ticks
+    plt.rcParams['ytick.labelsize'] = ticks
+    
+    fig, axs = plt.subplots(nrows=1, ncols=2, figsize=(27, 12))
+    
+    ax1, ax2 = axs[0], axs[1]
+    
+    # im0 = ax0.imshow(input_image, cmap=cmap,
+    #                  origin=origin)    
+    im1 = ax1.imshow(activation_1, cmap=cmap,
+                     origin=origin)
+    im2 = ax2.imshow(activation_2, cmap=cmap, 
+                     origin=origin) 
+    
+    ims = [im1, im2]
+    
+    for (i, ax) in enumerate(axs):
+        divider = make_axes_locatable(ax)
+        cax = divider.append_axes('right', size='5%', pad=0.05)
+        fig.colorbar(ims[i], cax=cax, orientation='vertical')
+        
+    # ax0.set_title('Input', fontsize=titles)
+    ax1.set_title('Activation 1', fontsize=titles)
+    ax2.set_title('Activation 2', fontsize=titles)
+    
+    return fig
+
 
 def predict_and_analyze(model_name, num_channels, dim, image):
     
@@ -130,31 +161,12 @@ def predict_and_analyze(model_name, num_channels, dim, image):
     
     origin = 'lower'
     
-    ##### Make the activation figure ######
-    plt.rcParams['xtick.labelsize'] = ticks
-    plt.rcParams['ytick.labelsize'] = ticks
+    # plot mean subtracted activations
+    fig1 = plot_activations(activation_1, activation_2, origin=origin)
     
-    fig, axs = plt.subplots(nrows=1, ncols=2, figsize=(27, 12))
-    
-    ax1, ax2 = axs[0], axs[1]
-    
-    # im0 = ax0.imshow(input_image, cmap=cmap,
-    #                  origin=origin)    
-    im1 = ax1.imshow(activation_1, cmap=cmap,
-                     origin=origin)
-    im2 = ax2.imshow(activation_2, cmap=cmap, 
-                     origin=origin) 
-    
-    ims = [im1, im2]
-    
-    for (i, ax) in enumerate(axs):
-        divider = make_axes_locatable(ax)
-        cax = divider.append_axes('right', size='5%', pad=0.05)
-        fig.colorbar(ims[i], cax=cax, orientation='vertical')
-        
-    # ax0.set_title('Input', fontsize=titles)
-    ax1.set_title('Activation 1', fontsize=titles)
-    ax2.set_title('Activation 2', fontsize=titles)
+    # plot raw activations
+    output, input_image, activation_1, activation_2 = get_activations(model, image, sub_mean=False)
+    fig2 = plot_activations(activation_1, activation_2, origin=origin)
     
     
     ##### make the figure for the input image #####
@@ -168,13 +180,13 @@ def predict_and_analyze(model_name, num_channels, dim, image):
     
     divider = make_axes_locatable(ax)
     cax = divider.append_axes('right', size='5%', pad=0.05)
-    fig.colorbar(im0, cax=cax, orientation='vertical')
+    input_fig.colorbar(im0, cax=cax, orientation='vertical')
         
     ax.set_title('Input', fontsize=titles)
     
     print("Sending to Hugging Face")
     
-    return output, input_fig, fig
+    return output, input_fig, fig1, fig2
 
 
 if __name__ == "__main__":
@@ -199,7 +211,8 @@ if __name__ == "__main__":
                 gr.Plot(label="Input Image", show_label=True), 
                 #  gr.Image(label="Activation 1", show_label=True), 
                 #  gr.Image(label="Actication 2", show_label=True)],
-                gr.Plot(label="Activations", show_label=True) 
+                gr.Plot(label="Mean-Subtracted Activations", show_label=True), 
+                gr.Plot(label="Raw Activations", show_label=True) 
                 ],
         title="Kinematic Planet Detector"
     )