Figure saving with proper size

Filter_Analysis/fgsm.py

@@ -121,7 +121,7 @@ def test(model, device, test_loader, epsilon):
             strength = 2*i + 1
 
             # Apply the filter with the specified strength
-            filtered_input = defense_filters.bilateral_filter(perturbed_data_normalized, batch_size=len(perturbed_data_normalized), d=strength)
+            filtered_input = defense_filters.gaussian_blur(perturbed_data_normalized, batch_size=len(perturbed_data_normalized), ksize=(strength, strength))
             # Evaluate the model on the filtered images
             filtered_output = model(filtered_input)
             # Get the predicted classification
@@ -158,7 +158,7 @@ def test(model, device, test_loader, epsilon):
     print(f"Unfiltered Accuracy = {unfiltered_correct} / {len(test_loader)} = {unfiltered_acc}")
     for i in range(TESTED_STRENGTH_COUNT):
         strength = 2*i + 1
-        print(f"Bilateral Filter (strength = {strength}) = {filtered_correct_counts[i]} / {len(test_loader)} = {filtered_accuracies[i]}")
+        print(f"Gaussian Blur (strength = {strength}) = {filtered_correct_counts[i]} / {len(test_loader)} = {filtered_accuracies[i]}")
 
     return unfiltered_acc, filtered_accuracies
 
@@ -172,14 +172,14 @@ for eps in epsilons:
     filtered_accuracies.append(filtered_accuracy)
 
 # Plot the results
+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"Bilateral Filter (strength = {2*i + 1})")
+    plt.plot(epsilons, filtered_accuracy, label=f"Gaussian Blur (strength = {2*i + 1})")
 
-plt.legend()
+plt.legend(loc="upper right")
-plt.title("Bilateral Filter Performance")
+plt.title("Gaussian Blur Performance")
 plt.xlabel("Attack Strength ($\\epsilon$)")
 plt.ylabel("Accuracy")
-
+plt.savefig("Images/GaussianBlurPerformance.png", )
-plt.show()

Filter_Analysis/wiki/Approach.md

@@ -2,8 +2,6 @@
 
 Attacking classifier models essentially boils down to adding precisely calculated noise to the input image, thereby tricking the classifier into selecting an incorrect class. The goal is to understand the efficacy of an array of denoising algorithms as adversarial machine learning defenses.
 
-## Individual Denoising Algorithms
+## Requirements
-
+For a given filter to be beneficial to th e
-## An Ensemble Approach
+1. The filter 
-
-## Training the Model on Filtered Data

Filter_Analysis/wiki/Results.md

@@ -954,3 +954,212 @@ Bilateral Filter (strength = 7) = 1999 / 10000 = 0.1999
 
 Bilateral Filter (strength = 9) = 1444 / 10000 = 0.1444
 
+
+## Gaussian Blur 
+====== EPSILON: 0.0 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 9920 / 10000 = 0.992
+
+Gaussian Blur (strength = 1) = 9920 / 10000 = 0.992
+
+Gaussian Blur (strength = 3) = 9879 / 10000 = 0.9879
+
+Gaussian Blur (strength = 5) = 9682 / 10000 = 0.9682
+
+Gaussian Blur (strength = 7) = 7731 / 10000 = 0.7731
+
+Gaussian Blur (strength = 9) = 5250 / 10000 = 0.525
+
+====== EPSILON: 0.025 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 9796 / 10000 = 0.9796
+
+Gaussian Blur (strength = 1) = 9796 / 10000 = 0.9796
+
+Gaussian Blur (strength = 3) = 9801 / 10000 = 0.9801
+
+Gaussian Blur (strength = 5) = 9512 / 10000 = 0.9512
+
+Gaussian Blur (strength = 7) = 7381 / 10000 = 0.7381
+
+Gaussian Blur (strength = 9) = 4862 / 10000 = 0.4862
+
+====== EPSILON: 0.05 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 9600 / 10000 = 0.96
+
+Gaussian Blur (strength = 1) = 9600 / 10000 = 0.96
+
+Gaussian Blur (strength = 3) = 9674 / 10000 = 0.9674
+
+Gaussian Blur (strength = 5) = 9271 / 10000 = 0.9271
+
+Gaussian Blur (strength = 7) = 6922 / 10000 = 0.6922
+
+Gaussian Blur (strength = 9) = 4446 / 10000 = 0.4446
+
+====== EPSILON: 0.07500000000000001 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 9260 / 10000 = 0.926
+
+Gaussian Blur (strength = 1) = 9260 / 10000 = 0.926
+
+Gaussian Blur (strength = 3) = 9460 / 10000 = 0.946
+
+Gaussian Blur (strength = 5) = 8939 / 10000 = 0.8939
+
+Gaussian Blur (strength = 7) = 6427 / 10000 = 0.6427
+
+Gaussian Blur (strength = 9) = 3989 / 10000 = 0.3989
+
+====== EPSILON: 0.1 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 8753 / 10000 = 0.8753
+
+Gaussian Blur (strength = 1) = 8753 / 10000 = 0.8753
+
+Gaussian Blur (strength = 3) = 9133 / 10000 = 0.9133
+
+Gaussian Blur (strength = 5) = 8516 / 10000 = 0.8516
+
+Gaussian Blur (strength = 7) = 5881 / 10000 = 0.5881
+
+Gaussian Blur (strength = 9) = 3603 / 10000 = 0.3603
+
+====== EPSILON: 0.125 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 8104 / 10000 = 0.8104
+
+Gaussian Blur (strength = 1) = 8104 / 10000 = 0.8104
+
+Gaussian Blur (strength = 3) = 8690 / 10000 = 0.869
+
+Gaussian Blur (strength = 5) = 7989 / 10000 = 0.7989
+
+Gaussian Blur (strength = 7) = 5278 / 10000 = 0.5278
+
+Gaussian Blur (strength = 9) = 3263 / 10000 = 0.3263
+
+====== EPSILON: 0.15000000000000002 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 7229 / 10000 = 0.7229
+
+Gaussian Blur (strength = 1) = 7229 / 10000 = 0.7229
+
+Gaussian Blur (strength = 3) = 8135 / 10000 = 0.8135
+
+Gaussian Blur (strength = 5) = 7415 / 10000 = 0.7415
+
+Gaussian Blur (strength = 7) = 4710 / 10000 = 0.471
+
+Gaussian Blur (strength = 9) = 2968 / 10000 = 0.2968
+
+====== EPSILON: 0.17500000000000002 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 6207 / 10000 = 0.6207
+
+Gaussian Blur (strength = 1) = 6207 / 10000 = 0.6207
+
+Gaussian Blur (strength = 3) = 7456 / 10000 = 0.7456
+
+Gaussian Blur (strength = 5) = 6741 / 10000 = 0.6741
+
+Gaussian Blur (strength = 7) = 4224 / 10000 = 0.4224
+
+Gaussian Blur (strength = 9) = 2683 / 10000 = 0.2683
+
+====== EPSILON: 0.2 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 5008 / 10000 = 0.5008
+
+Gaussian Blur (strength = 1) = 5008 / 10000 = 0.5008
+
+Gaussian Blur (strength = 3) = 6636 / 10000 = 0.6636
+
+Gaussian Blur (strength = 5) = 5983 / 10000 = 0.5983
+
+Gaussian Blur (strength = 7) = 3755 / 10000 = 0.3755
+
+Gaussian Blur (strength = 9) = 2453 / 10000 = 0.2453
+
+====== EPSILON: 0.225 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 3894 / 10000 = 0.3894
+
+Gaussian Blur (strength = 1) = 3894 / 10000 = 0.3894
+
+Gaussian Blur (strength = 3) = 5821 / 10000 = 0.5821
+
+Gaussian Blur (strength = 5) = 5243 / 10000 = 0.5243
+
+Gaussian Blur (strength = 7) = 3359 / 10000 = 0.3359
+
+Gaussian Blur (strength = 9) = 2269 / 10000 = 0.2269
+
+====== EPSILON: 0.25 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 2922 / 10000 = 0.2922
+
+Gaussian Blur (strength = 1) = 2922 / 10000 = 0.2922
+
+Gaussian Blur (strength = 3) = 5050 / 10000 = 0.505
+
+Gaussian Blur (strength = 5) = 4591 / 10000 = 0.4591
+
+Gaussian Blur (strength = 7) = 3034 / 10000 = 0.3034
+
+Gaussian Blur (strength = 9) = 2112 / 10000 = 0.2112
+
+====== EPSILON: 0.275 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 2149 / 10000 = 0.2149
+
+Gaussian Blur (strength = 1) = 2149 / 10000 = 0.2149
+
+Gaussian Blur (strength = 3) = 4290 / 10000 = 0.429
+
+Gaussian Blur (strength = 5) = 3998 / 10000 = 0.3998
+
+Gaussian Blur (strength = 7) = 2743 / 10000 = 0.2743
+
+Gaussian Blur (strength = 9) = 1983 / 10000 = 0.1983
+
+====== EPSILON: 0.30000000000000004 ======
+
+Clean (No Filter) Accuracy = 9920 / 10000 = 0.992
+
+Unfiltered Accuracy = 1599 / 10000 = 0.1599
+
+Gaussian Blur (strength = 1) = 1599 / 10000 = 0.1599
+
+Gaussian Blur (strength = 3) = 3648 / 10000 = 0.3648
+
+Gaussian Blur (strength = 5) = 3481 / 10000 = 0.3481
+
+Gaussian Blur (strength = 7) = 2493 / 10000 = 0.2493
+
+Gaussian Blur (strength = 9) = 1884 / 10000 = 0.1884