normalisation for patch-infer fixed

app.py

@@ -48,8 +48,18 @@ def infer_full_vol(tensor, model):
 
 def infer_patch_based(tensor, model, patch_size=64, stride_length=32, stride_width=32, stride_depth=16, batch_size=10, num_worker=2):
     test_subject = tio.Subject(img = tio.ScalarImage(tensor=tensor.unsqueeze(0))) # adding channel dim while creating the TorchIO subject
+                
     overlap = np.subtract(patch_size, (stride_length, stride_width, stride_depth))
     
+    def normaliser(batch):
+        """
+        Purpose: Normalise pixel intensities of each patch using the max values in the 3D patch
+        :param batch: 5D array (batch_size x channel x width x depth x height)
+        """
+        for i in range(batch.shape[0]):
+            batch[i] = batch[i] / batch[i].max()
+        return batch
+    
     with torch.no_grad():
         grid_sampler = tio.inference.GridSampler(
                             test_subject,
@@ -63,7 +73,7 @@ def infer_patch_based(tensor, model, patch_size=64, stride_length=32, stride_wid
         for i, patches_batch in enumerate(patch_loader):
             st.text(f"Processing batch {i + 1} of {total_batches}...")
             
-            local_batch = patches_batch['img'][tio.DATA].float()
+            local_batch = normaliser(patches_batch['img'][tio.DATA].float())
             local_batch = local_batch / local_batch.max()
             locations = patches_batch[tio.LOCATION]