diff --git a/ganimation/attacks.py b/ganimation/attacks.py
index 3089a2f..ca3be29 100644
--- a/ganimation/attacks.py
+++ b/ganimation/attacks.py
@@ -80,8 +80,7 @@ class LinfPGDAttack(object):
 
     def perturb_iter_class(self, X_nat, y, c_trg):
         """
-        Given examples (X_nat, y), returns adversarial
-        examples within epsilon of X_nat in l_infinity norm.
+        Iterative Class Conditional Attack
         """
         X = X_nat.clone().detach_()
 
@@ -113,6 +112,38 @@ class LinfPGDAttack(object):
 
         return X, eta
 
+    def perturb_joint_class(self, X_nat, y, c_trg):
+        """
+        Joint Class Conditional Attack
+        """
+        X = X_nat.clone().detach_()
+
+        J = c_trg.size(0)
+        full_loss = 0.0
+
+        for i in range(self.k):
+            for j in range(J):
+                # print(i)
+                X.requires_grad = True
+                output_att, output_img = self.model(X, c_trg[j,:].unsqueeze(0))
+
+                out = imFromAttReg(output_att, output_img, X)
+
+                self.model.zero_grad()
+
+                loss = -self.loss_fn(output_att, y)
+                full_loss += loss
+
+            full_loss.backward()
+            grad = X.grad
+
+            X_adv = X + self.a * grad.sign()
+
+            eta = torch.clamp(X_adv - X_nat, min=-self.epsilon, max=self.epsilon)
+            X = torch.clamp(X_nat + eta, min=-1, max=1).detach_()
+
+        return X, eta
+
 def clip_tensor(X, Y, Z):
     # Clip X with Y min and Z max
     X_np = X.data.cpu().numpy()
diff --git a/ganimation/solver.py b/ganimation/solver.py
index 310426f..0a5c0c3 100644
--- a/ganimation/solver.py
+++ b/ganimation/solver.py
@@ -404,29 +404,47 @@ class Solver(Utils):
             x_advs = []
 
             for idx, image_path in enumerate(images_to_animate_path):
-                image_to_animate = regular_image_transform(
-                    Image.open(image_path)).unsqueeze(0).cuda()
+                image_to_animate = regular_image_transform(Image.open(image_path)).unsqueeze(0).cuda()
 
                 all_images = torch.cat([regular_image_transform(Image.open(path)).unsqueeze(0) for path in images_to_animate_path], dim=0).cuda()
 
-                if idx == 0:
-                    for target_idx in range(targets.size(0)):
-                        x_adv, perturb = pgd_attack.perturb(image_to_animate, black, targets[target_idx, :].unsqueeze(0).cuda())
-                        x_advs.append((x_adv, perturb))
+                # Transfer to different images
+                # if idx == 0:
+                #     for target_idx in range(targets.size(0)):
+                #         x_adv, perturb = pgd_attack.perturb(image_to_animate, black, targets[target_idx, :].unsqueeze(0).cuda())
+                #         x_advs.append((x_adv, perturb))
 
                 for target_idx in range(targets.size(0)):
-                    # if target_idx == 0:
-                    #     img = regular_image_transform(Image.open(images_to_animate_path[idx])).unsqueeze(0).cuda()
-                    #     # x_adv, perturb = pgd_attack.perturb(img, black, targets[0, :].unsqueeze(0).cuda())
-                    #     x_adv, perturb = pgd_attack.perturb_iter_class(image_to_animate, black, targets[:, :].cuda())
-                    #     # _, perturb = pgd_attack.perturb_iter_data(image_to_animate, all_images, black, targets[68, :].unsqueeze(0).cuda())
+                    # Transfer to different classes
+                    if target_idx == 0:
+                        # img = regular_image_transform(Image.open(images_to_animate_path[idx])).unsqueeze(0).cuda()
+
+                        # Wrong Class
+                        x_adv, perturb = pgd_attack.perturb(image_to_animate, black, targets[0, :].unsqueeze(0).cuda()
+
+                        # Joint Class Conditional
+                        # x_adv, perturb = pgd_attack.perturb_joint_class(image_to_animate, black, targets[:, :].cuda())
+
+                        # Iterative Class Conditional
+                        # x_adv, perturb = pgd_attack.perturb_iter_class(image_to_animate, black, targets[:, :].cuda())
+                        
+                        # Iterative Data
+                        # _, perturb = pgd_attack.perturb_iter_data(image_to_animate, all_images, black, targets[68, :].unsqueeze(0).cuda())
                         
                     targets_au = targets[target_idx, :].unsqueeze(0).cuda()
+
+                    # Normal Attack
                     # x_adv, perturb = pgd_attack.perturb(image_to_animate, black, targets_au)
-                    x_adv, perturb = x_advs[target_idx]
-                    # x_adv = image_to_animate + perturb
-                    x_adv = image_to_animate
+
+                    # x_adv, perturb = x_advs[target_idx]
+
+                    x_adv = image_to_animate + perturb
+
+                    # No Attack
+                    # x_adv = image_to_animate
+
                     # print(image_to_animate.shape, x_adv.shape)
+
                     with torch.no_grad():
                         resulting_images_att, resulting_images_reg = self.G(
                             x_adv, targets_au)