default images show on app startup

Signed-off-by: Kieran Fraser <Kieran.Fraser@ibm.com>

app.py

@@ -192,7 +192,54 @@ def show_params(type):
         return gr.Column(visible=True)
     return gr.Column(visible=False)  
 
+def default_clean():
+    return [('./data/pgd/clean/0_airplane.png', 'airplane'),
+            ('./data/pgd/clean/1_automobile.png', 'automobile'),
+            ('./data/pgd/clean/2_bird.png', 'bird'),
+            ('./data/pgd/clean/3_cat.png', 'cat'),
+            ('./data/pgd/clean/4_deer.png', 'deer'),
+            ('./data/pgd/clean/5_dog.png', 'dog'),
+            ('./data/pgd/clean/6_frog.png', 'frog'),
+            ('./data/pgd/clean/7_horse.png', 'horse'),
+            ('./data/pgd/clean/8_ship.png', 'ship'),
+            ('./data/pgd/clean/9_truck.png', 'truck')]
+    
+def default_perturbation():
+    return [('./data/pgd/perturb/p1.png'),
+            ('./data/pgd/perturb/p2.png'),
+            ('./data/pgd/perturb/p3.png'),
+            ('./data/pgd/perturb/p4.png'),
+            ('./data/pgd/perturb/p5.png'),
+            ('./data/pgd/perturb/p6.png'),
+            ('./data/pgd/perturb/p7.png'),
+            ('./data/pgd/perturb/p8.png'),
+            ('./data/pgd/perturb/p9.png'),
+            ('./data/pgd/perturb/p10.png')]
+    
+def default_pgd():
+    return [('./data/pgd/attacked/0_airplane.png', 'airplane'),
+            ('./data/pgd/attacked/1_automobile.png', 'automobile'),
+            ('./data/pgd/attacked/2_bird.png', 'bird'),
+            ('./data/pgd/attacked/3_cat.png', 'cat'),
+            ('./data/pgd/attacked/4_deer.png', 'deer'),
+            ('./data/pgd/attacked/5_dog.png', 'dog'),
+            ('./data/pgd/attacked/6_frog.png', 'frog'),
+            ('./data/pgd/attacked/7_horse.png', 'horse'),
+            ('./data/pgd/attacked/8_ship.png', 'ship'),
+            ('./data/pgd/attacked/9_truck.png', 'truck')]
 
+def default_patch():
+    return [('./data/patch/0_airplane.png', 'airplane'),
+            ('./data/patch/1_automobile.png', 'automobile'),
+            ('./data/patch/2_bird.png', 'bird'),
+            ('./data/patch/3_cat.png', 'cat'),
+            ('./data/patch/4_deer.png', 'deer'),
+            ('./data/patch/5_dog.png', 'dog'),
+            ('./data/patch/6_frog.png', 'frog'),
+            ('./data/patch/7_horse.png', 'horse'),
+            ('./data/patch/8_ship.png', 'ship'),
+            ('./data/patch/9_truck.png', 'truck')]
+    
 # e.g. To use a local alternative theme: carbon_theme = Carbon()
 carbon_theme = Carbon()
 with gr.Blocks(css=css, theme='Tshackelton/IBMPlex-DenseReadable') as demo:
@@ -224,9 +271,9 @@ with gr.Blocks(css=css, theme='Tshackelton/IBMPlex-DenseReadable') as demo:
     
     with gr.Row(elem_classes=["larger-gap", "custom-text"]):
         with gr.Column(scale=1, elem_classes="cust-width"):
-            gr.Markdown('''<p style="font-size: 20px; text-align: justify">ℹ️ First lets set the scene. You have a dataset of images, such as CIFAR-10.</p>''')
+            gr.Markdown('''<p style="font-size: 20px; text-align: justify">ℹ️  First lets set the scene. You have a dataset of images, such as CIFAR-10.</p>''')
             gr.Markdown('''<p style="font-size: 18px; text-align: justify"><i>Note: CIFAR-10 images are low resolution images which span 10 different categories as shown.</i></p>''')
-            gr.Markdown('''<p style="font-size: 20px; text-align: justify">ℹ️ Your goal is to have an AI model capable of classifying these images. So you 
+            gr.Markdown('''<p style="font-size: 20px; text-align: justify">ℹ️  Your goal is to have an AI model capable of classifying these images. So you 
                         train a model on this dataset, or use a pre-trained model from Hugging Face,
                         such as Meta's Distilled Data-efficient Image Transformer.</p>''')
         with gr.Column(scale=1, elem_classes="cust-width"): 
@@ -248,11 +295,11 @@ with gr.Blocks(css=css, theme='Tshackelton/IBMPlex-DenseReadable') as demo:
             
     gr.Markdown('''<hr/>''')
     
-    gr.Markdown('''<p style="text-align: justify; font-size: 18px">ℹ️ Now as a responsible AI expert, you wish to assert that your model is not vulnerable to
+    gr.Markdown('''<p style="text-align: justify; font-size: 18px">ℹ️  Now as a responsible AI expert, you wish to assert that your model is not vulnerable to
                 attacks which might manipulate the prediction. For instance, ships become classified as birds. To do this, you will deploy
                 adversarial attacks against your own model and assess its performance.</p>''')
     
-    gr.Markdown('''<p style="text-align: justify; font-size: 18px">ℹ️ Below are two common types of evasion attack. Both create adversarial images, which at first glance, seem the same as the original images,
+    gr.Markdown('''<p style="text-align: justify; font-size: 18px">ℹ️  Below are two common types of evasion attack. Both create adversarial images, which at first glance, seem the same as the original images,
                 however they contain subtle changes which cause the AI model to make incorrect predictions.</p><br/>''')
 
         
@@ -276,25 +323,26 @@ with gr.Blocks(css=css, theme='Tshackelton/IBMPlex-DenseReadable') as demo:
                 with gr.Row(elem_classes='symbols'):
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>The unmodified, original CIFAR-10 images, with model predictions.</i></p><br>''')
-                        original_gallery = gr.Gallery(label="Original", preview=False, show_download_button=True)
+                        original_gallery = gr.Gallery(default_clean, label="Original", preview=False, show_download_button=True)
                         benign_output = gr.Label(num_top_classes=3, visible=False)
-                        clean_accuracy = gr.Number(label="Clean Accuracy", precision=2)
+                        clean_accuracy = gr.Number(1, label="Clean Accuracy", precision=2)
                     with gr.Column(scale=1, min_width=0, elem_classes='symbols'):
                         gr.Markdown('''➕''')
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>Visual representation of the calculated perturbations for attacking the model.</i></p><br>''')
-                        delta_gallery = gr.Gallery(label="Added perturbation", preview=False, show_download_button=True)
+                        delta_gallery = gr.Gallery(default_perturbation, label="Added perturbation", preview=False, show_download_button=True)
                     with gr.Column(scale=1, min_width=0):
                         gr.Markdown('''🟰''', elem_classes='symbols')
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>The original image (with optimized perturbations applied) gives us an adversarial image which fools the model.</i></p>''')
-                        adversarial_gallery = gr.Gallery(label="Adversarial", preview=False, show_download_button=True)
+                        adversarial_gallery = gr.Gallery(default_pgd, label="Adversarial", preview=False, show_download_button=True)
                         adversarial_output = gr.Label(num_top_classes=3, visible=False)
-                        robust_accuracy = gr.Number(label="Robust Accuracy", precision=2)
+                        robust_accuracy = gr.Number(0, label="Robust Accuracy", precision=2)
                         
             bt_eval_pgd.click(clf_evasion_evaluate, inputs=[attack, max_iter, eps, eps_steps, attack, attack, attack, attack],
                                                     outputs=[original_gallery, adversarial_gallery, delta_gallery, clean_accuracy,
                                                             robust_accuracy])
+                
     
     gr.Markdown('''<br/>''')
       
@@ -319,23 +367,23 @@ with gr.Blocks(css=css, theme='Tshackelton/IBMPlex-DenseReadable') as demo:
                 with gr.Row(elem_classes='symbols'):
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>The unmodified, original CIFAR-10 images, with model predictions.</i></p><br><br>''')
-                        original_gallery = gr.Gallery(label="Original", preview=False, show_download_button=True)
-                        clean_accuracy = gr.Number(label="Clean Accuracy", precision=2)
+                        original_gallery = gr.Gallery(default_clean, label="Original", preview=False, show_download_button=True)
+                        clean_accuracy = gr.Number(1, label="Clean Accuracy", precision=2)
                     
                     with gr.Column(scale=1, min_width=0, elem_classes='symbols'):
                         gr.Markdown('''➕''')
                         
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>Visual representation of the optimized patch for attacking the model.</i></p><br><br>''')
-                        delta_gallery = gr.Gallery(label="Patches", preview=True, show_download_button=True)
+                        delta_gallery = gr.Gallery(['./data/patch/patch.png'], label="Patches", preview=True, show_download_button=True)
                         
                     with gr.Column(scale=1, min_width=0):
                         gr.Markdown('''🟰''', elem_classes='symbols')
                         
                     with gr.Column(scale=10):
                         gr.Markdown('''<p style="font-size: 18px"><i>The original image (with optimized perturbations applied) gives us an adversarial image which fools the model.</i></p>''')
-                        adversarial_gallery = gr.Gallery(label="Adversarial", preview=False, show_download_button=True)
-                        robust_accuracy = gr.Number(label="Robust Accuracy", precision=2)
+                        adversarial_gallery = gr.Gallery(default_patch, label="Adversarial", preview=False, show_download_button=True)
+                        robust_accuracy = gr.Number(0.8, label="Robust Accuracy", precision=2)
                         
             eval_btn_patch.click(clf_evasion_evaluate, inputs=[attack, max_iter, eps, eps_steps, x_location, y_location, patch_height,
                                                                 patch_width],