signal-grad-cam 0.1.5__tar.gz → 0.1.7__tar.gz

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: signal_grad_cam
-Version: 0.1.5
+Version: 0.1.7
 Summary: SignalGrad-CAM aims at generalising Grad-CAM to one-dimensional applications, while enhancing usability and efficiency.
 Home-page: https://github.com/samuelepe11/signal_grad_cam
 Author: Samuele Pe

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r") as f:
 
 setup(
     name="signal_grad_cam",
-    version="0.1.5",
+    version="0.1.7",
     description="SignalGrad-CAM aims at generalising Grad-CAM to one-dimensional applications, while enhancing usability"
                 " and efficiency.",
     keywords="XAI, class activation maps, CNN, time series",

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/signal_grad_cam/cam_builder.py

@@ -92,15 +92,20 @@ class CamBuilder:
         # Show available 1D or 2D convolutional layers
         print()
         print("SEARCHING FOR NETWORK LAYERS:")
-        self.__print_justify("Please, verify that your network contains at least one 1D or 2D convolutional layer "
-                             "and note the names of the layers that are of interest to you. If the desired layer is not"
-                             " present in the following list, it can still be accessed by the name used in the network "
-                             "to identify it.\nVerify whether the model ends with an activation function from the "
-                             "Softmax family (such as Sigmoid o Softmax). Even if this activation function is not "
-                             "present in the following list, ensure to check if it is applied at the end of the "
-                             "network. Please make sure that the provided model is set in inference ('eval') mode for "
-                             "PyTorch models and that TensorFlow/Keras models have been built (they must have the "
-                             "specific 'inputs' and 'output' attributes)\nNetwork layers found (name: type)")
+        self.__print_justify("Please verify that your network contains at least one 1D or 2D convolutional layer, "
+                             "and take note of the names of the layers that are of interest to you. If the desired "
+                             "layer is not present in the list below, it can still be accessed using the name by which "
+                             "it is defined in the network.\n"
+                             "Also, check whether the model ends with an activation function from the Softmax family "
+                             "(such as Sigmoid or Softmax). Even if this activation function is not listed below, you "
+                             "must indicate its presence using the appropriate argument in the 'get_cam' function. "
+                             "Note that in binary classification networks (those ending with a Sigmoid function), "
+                             "overconfident predictions can cause the Sigmoid to saturate, leading to empty or null "
+                             "maps. To prevent this, modify your network to output logits directly.\n"
+                             "Make sure the provided model is set to inference mode ('eval') if using PyTorch. For "
+                             "TensorFlow/Keras models, ensure the model is built—i.e., it must have defined 'inputs' "
+                             "and 'output' attributes.\n"
+                             "Network layers found (name: type):")
         self._get_layers_pool(show=True, extend_search=extend_search)
         print()
 
@@ -578,8 +583,8 @@ class CamBuilder:
             data_shape_list_processed = [tuple(data_element.shape) for data_element in data_list]
             if len(set(data_shape_list_processed)) != 1:
                 data_list = [np.resize(x, data_shape_list_processed[0]) for x in data_list]
-                self.__print_justify("Input data items have different shapes. Each item has been reshaped to match the "
-                                     "first item's dimensions for batching. To prevent this, provide one item at a "
+                self.__print_justify("\nInput data items have different shapes. Each item has been reshaped to match the"
+                                     " first item's dimensions for batching. To prevent this, provide one item at a "
                                      "time.")
 
         cam_list, target_probs = self._create_raw_batched_cams(data_list, target_class, target_layer, explainer_type,

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/signal_grad_cam/pytorch_cam_builder.py

@@ -207,12 +207,13 @@ class TorchCamBuilder(CamBuilder):
             if len(outputs.shape) == 2 and outputs.shape[1] > 1:
                 target_probs = torch.softmax(target_scores, dim=1)
             else:
+                p = torch.sigmoid(outputs)
                 if len(outputs.shape) == 1:
                     target_scores = torch.stack([-outputs, outputs], dim=1)
+                    target_probs = torch.stack([1 - p, p], dim=1)
                 elif len(outputs.shape) == 2 and outputs.shape[1] == 1:
                     target_scores = torch.cat([-outputs, outputs], dim=1)
-                p = torch.sigmoid(outputs)
-                target_probs = torch.stack([1 - p, p], dim=1)
+                    target_probs = torch.cat([1 - p, p], dim=1)
 
         target_probs = target_probs[:, target_class].cpu().detach().numpy()
 

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/signal_grad_cam/tensorflow_cam_builder.py

@@ -222,12 +222,13 @@ class TfCamBuilder(CamBuilder):
                 if len(outputs.shape) == 2 and outputs.shape[1] > 1:
                     target_probs = tf.nn.softmax(target_scores, axis=1)
                 else:
+                    p = tf.math.sigmoid(outputs)
                     if len(outputs.shape) == 1:
                         target_scores = tf.stack([-outputs, outputs], axis=1)
+                        target_probs = tf.stack([1 - p, p], axis=1)
                     elif len(outputs.shape) == 2 and outputs.shape[1] == 1:
                         target_scores = tf.concat([-outputs, outputs], axis=1)
-                    p = tf.math.sigmoid(outputs)
-                    target_probs = tf.stack([1 - p, p], axis=1)
+                        target_probs = tf.concat([1 - p, p], axis=1)
 
             target_scores = target_scores[:, target_class]
             target_probs = target_probs[:, target_class]

{signal_grad_cam-0.1.5 → signal_grad_cam-0.1.7}/signal_grad_cam.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: signal-grad-cam
-Version: 0.1.5
+Version: 0.1.7
 Summary: SignalGrad-CAM aims at generalising Grad-CAM to one-dimensional applications, while enhancing usability and efficiency.
 Home-page: https://github.com/samuelepe11/signal_grad_cam
 Author: Samuele Pe