diff --git a/run_ssd_example_output.jpg b/run_ssd_example_output.jpg
deleted file mode 100644
index 3075c9b9..00000000
Binary files a/run_ssd_example_output.jpg and /dev/null differ
diff --git a/run_ssd_live_caffe2.py b/run_ssd_live_caffe2.py
index b1fa40fa..526a2802 100644
--- a/run_ssd_live_caffe2.py
+++ b/run_ssd_live_caffe2.py
@@ -83,12 +83,13 @@ def predict(width, height, confidences, boxes, prob_threshold, iou_threshold=0.5
     interval = timer.end()
     print('Inference Time: {:.2f}s.'.format(interval))
     timer.start()
-    boxes, labels, probs = predict(orig_image.shape[1], orig_image.shape[0], confidences, boxes, 0.4)
+    boxes, labels, probs = predict(orig_image.shape[1], orig_image.shape[0], confidences, boxes, 0.55)
     interval = timer.end()
     print('NMS Time: {:.2f}s, Detect Objects: {:d}.'.format(interval, labels.shape[0]))
     for i in range(boxes.shape[0]):
         box = boxes[i, :]
-        label = class_names[labels[i]]
+        label = f"{class_names[labels[i]]}: {probs[i]:.2f}"
+
         cv2.rectangle(orig_image, (box[0], box[1]), (box[2], box[3]), (255, 255, 0), 4)
 
         cv2.putText(orig_image, label,
diff --git a/train_ssd.py b/train_ssd.py
index 6820fbff..43f6a050 100644
--- a/train_ssd.py
+++ b/train_ssd.py
@@ -13,6 +13,7 @@
 from vision.ssd.vgg_ssd import create_vgg_ssd
 from vision.ssd.mobilenetv1_ssd import create_mobilenetv1_ssd
 from vision.ssd.mobilenetv1_ssd_lite import create_mobilenetv1_ssd_lite
+from vision.ssd.mobilenet_v2_ssd_lite import create_mobilenetv2_ssd_lite
 from vision.ssd.squeezenet_ssd_lite import create_squeezenet_ssd_lite
 from vision.datasets.voc_dataset import VOCDataset
 from vision.datasets.open_images import OpenImagesDataset
@@ -35,12 +36,15 @@
 
 
 parser.add_argument('--net', default="vgg16-ssd",
-                    help="The network architecture, it can be mb1-ssd, mb1-lite-ssd or vgg16-ssd.")
+                    help="The network architecture, it can be mb1-ssd, mb1-lite-ssd, mb2-ssd-lite or vgg16-ssd.")
 parser.add_argument('--freeze_base_net', action='store_true',
                     help="Freeze base net layers.")
 parser.add_argument('--freeze_net', action='store_true',
                     help="Freeze all the layers except the prediction head.")
 
+parser.add_argument('--mb2_width_mult', default=1.0, type=float,
+                    help='Width Multiplifier for MobilenetV2')
+
 # Params for SGD
 parser.add_argument('--lr', '--learning-rate', default=1e-3, type=float,
                     help='initial learning rate')
@@ -179,11 +183,13 @@ def test(loader, net, criterion, device):
     elif args.net == 'sq-ssd-lite':
         create_net = create_squeezenet_ssd_lite
         config = squeezenet_ssd_config
+    elif args.net == 'mb2-ssd-lite':
+        create_net = lambda num: create_mobilenetv2_ssd_lite(num, width_mult=args.mb2_width_mult)
+        config = mobilenetv1_ssd_config
     else:
         logging.fatal("The net type is wrong.")
         parser.print_help(sys.stderr)
         sys.exit(1)
-
     train_transform = TrainAugmentation(config.image_size, config.image_mean, config.image_std)
     target_transform = MatchPrior(config.priors, config.center_variance,
                                   config.size_variance, 0.5)
diff --git a/vision/ssd/ssd.py b/vision/ssd/ssd.py
index d15ab75d..962b9a22 100644
--- a/vision/ssd/ssd.py
+++ b/vision/ssd/ssd.py
@@ -5,6 +5,8 @@
 import torch.nn.functional as F
 
 from ..utils import box_utils
+from collections import namedtuple
+GraphPath = namedtuple("GraphPath", ['s0', 'name', 's1'])  #
 
 
 class SSD(nn.Module):
@@ -25,7 +27,8 @@ def __init__(self, num_classes: int, base_net: nn.ModuleList, source_layer_index
         self.config = config
 
         # register layers in source_layer_indexes by adding them to a module list
-        self.source_layer_add_ons = nn.ModuleList([t[1] for t in source_layer_indexes if isinstance(t, tuple)])
+        self.source_layer_add_ons = nn.ModuleList([t[1] for t in source_layer_indexes
+                                                   if isinstance(t, tuple) and not isinstance(t, GraphPath)])
         if device:
             self.device = device
         else:
@@ -40,19 +43,32 @@ def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
         start_layer_index = 0
         header_index = 0
         for end_layer_index in self.source_layer_indexes:
-
-            if isinstance(end_layer_index, tuple):
+            if isinstance(end_layer_index, GraphPath):
+                path = end_layer_index
+                end_layer_index = end_layer_index.s0
+                added_layer = None
+            elif isinstance(end_layer_index, tuple):
                 added_layer = end_layer_index[1]
                 end_layer_index = end_layer_index[0]
+                path = None
             else:
                 added_layer = None
+                path = None
             for layer in self.base_net[start_layer_index: end_layer_index]:
                 x = layer(x)
-            start_layer_index = end_layer_index
             if added_layer:
                 y = added_layer(x)
             else:
                 y = x
+            if path:
+                sub = getattr(self.base_net[end_layer_index], path.name)
+                for layer in sub[:path.s1]:
+                    x = layer(x)
+                y = x
+                for layer in sub[path.s1:]:
+                    x = layer(x)
+                end_layer_index += 1
+            start_layer_index = end_layer_index
             confidence, location = self.compute_header(header_index, y)
             header_index += 1
             confidences.append(confidence)