Finished testing random noise and threshold filter

Filter_Analysis/defense_filters.py

@@ -4,7 +4,6 @@ import numpy as np
 import torch
 
 
-
 # Turn a pytorch tensor into an image
 def pttensor_to_images(data):
     images = None

Filter_Analysis/fgsm.py

@@ -118,10 +118,10 @@ def test(model, device, test_loader, epsilon):
 
         # Evaluate performance for 
         for i in range(TESTED_STRENGTH_COUNT):
-            strength = 2*i + 1
+            strength = (2*i + 1)/10
 
             # Apply the filter with the specified strength
-            filtered_input = defense_filters.gaussian_blur(perturbed_data_normalized, batch_size=len(perturbed_data_normalized), ksize=(strength, strength))
+            filtered_input = defense_filters.threshold_filter(perturbed_data_normalized, batch_size=len(perturbed_data_normalized), threshold=strength)
             # Evaluate the model on the filtered images
             filtered_output = model(filtered_input)
             # Get the predicted classification
@@ -157,8 +157,8 @@ def test(model, device, test_loader, epsilon):
     print(f"Clean (No Filter) Accuracy = {orig_correct} / {len(test_loader)} = {orig_acc}")
     print(f"Unfiltered Accuracy = {unfiltered_correct} / {len(test_loader)} = {unfiltered_acc}")
     for i in range(TESTED_STRENGTH_COUNT):
-        strength = 2*i + 1
-        print(f"Gaussian Blur (strength = {strength}) = {filtered_correct_counts[i]} / {len(test_loader)} = {filtered_accuracies[i]}")
+        strength = (2*i + 1)/10
+        print(f"Threshold Filter (threshold = {strength}) = {filtered_correct_counts[i]} / {len(test_loader)} = {filtered_accuracies[i]}")
 
     return unfiltered_acc, filtered_accuracies
 
@@ -176,10 +176,10 @@ plt.figure(figsize=(16,9))
 plt.plot(epsilons, unfiltered_accuracies, label="Attacked Accuracy")
 for i in range(TESTED_STRENGTH_COUNT):
     filtered_accuracy = [filter_eps[i] for filter_eps in filtered_accuracies]
-    plt.plot(epsilons, filtered_accuracy, label=f"Gaussian Blur (strength = {2*i + 1})")
+    plt.plot(epsilons, filtered_accuracy, label=f"Threshold Filter (threshold = {(2*i + 1)/10})")
 
 plt.legend(loc="upper right")
-plt.title("Gaussian Blur Performance")
+plt.title("Threshold Filter Performance")
 plt.xlabel("Attack Strength ($\\epsilon$)")
 plt.ylabel("Accuracy")
-plt.savefig("Images/GaussianBlurPerformance.png", )
+plt.savefig("Images/ThresholdFilterPerformance.png", )