Fix

Author: enezhadian

model/cnn_geometric_model.py

@@ -15,8 +15,6 @@
 
 def featureL2Norm(feature):
     epsilon = 1e-6
-    #        print(feature.size())
-    #        print(torch.pow(torch.sum(torch.pow(feature,2),1)+epsilon,0.5).size())
     norm = torch.pow(torch.sum(torch.pow(feature, 2), 1) + epsilon,
                      0.5).unsqueeze(1).expand_as(feature)
     return torch.div(feature, norm)
@@ -136,12 +134,7 @@ def __init__(self, shape='3D', normalization=True):
         self.shape = shape
         self.ReLU = nn.ReLU()
 
-    def forward(self, feature_A, feature_B, cues_A, cues_B):
-        if cues_A is None or cues_B is None:
-            # localization cues are available
-            pass # TODO: create uniform localization weight over all the images.
-
-        # TODO: update this to use localization cues.
+    def forward(self, feature_A, feature_B):
         b, c, h, w = feature_A.size()
         if self.shape == '3D':
             # reshape features for matrix multiplication
@@ -223,9 +216,6 @@ def __init__(
         use_cuda=True,
         delf_path=''
     ):
-        #                 regressor_channels_1 = 128,
-        #                 regressor_channels_2 = 64):
-
         super(CNNGeometric, self).__init__()
         self.use_cuda = use_cuda
         self.feature_self_matching = feature_self_matching
@@ -259,11 +249,8 @@ def forward(self, tnf_batch):
         # feature extraction
         feature_A = self.FeatureExtraction(tnf_batch['source_image'])
         feature_B = self.FeatureExtraction(tnf_batch['target_image'])
-        # localization cues
-        cues_A = tnf_batch.get('source_cues')
-        cues_B = tnf_batch.get('target_cues')
         # feature correlation
-        correlation = self.FeatureCorrelation(feature_A, feature_B, cues_A, cues_B)
+        correlation = self.FeatureCorrelation(feature_A, feature_B)
         # regression to tnf parameters theta
         theta = self.FeatureRegression(correlation)
 

model/delf.py

@@ -74,5 +74,5 @@ def forward(self, batch):
         interim = F.relu(self.bn1(interim))
         interim = self.conv2(interim)
         attention = F.softplus(interim)
-        weight = (attention - attention.min())/(attention.max() - attention.min())
-        return weight*feature_maps
+        weight = (attention - attention.min()) / (attention.max() - attention.min())
+        return weight * feature_maps

options/options.py

@@ -264,12 +264,7 @@ def add_cnn_model_parameters(self):
             default='',
             help='feature extraction CNN last layer'
         )
-        model_params.add_argument(
-            '--delf-path',
-            type=str,
-            default='',
-            help='DELF checkpoint path'
-        )
+        model_params.add_argument('--delf-path', type=str, default='', help='DELF checkpoint path')
         model_params.add_argument(
             '--fr-feature-size', type=int, default=15, help='image input size'
         )