diff --git a/torchvision/models/detection/backbone_utils.py b/torchvision/models/detection/backbone_utils.py
index d662de8078a..3178a81b52c 100644
--- a/torchvision/models/detection/backbone_utils.py
+++ b/torchvision/models/detection/backbone_utils.py
@@ -151,7 +151,7 @@ def mobilenet_backbone(
 
     # find the index of the layer from which we wont freeze
     assert 0 <= trainable_layers <= num_stages
-    freeze_before = num_stages if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
+    freeze_before = len(backbone) if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
 
     for b in backbone[:freeze_before]:
         for parameter in b.parameters():
diff --git a/torchvision/models/detection/ssd.py b/torchvision/models/detection/ssd.py
index 6695167d462..b6e620dba4a 100644
--- a/torchvision/models/detection/ssd.py
+++ b/torchvision/models/detection/ssd.py
@@ -510,7 +510,7 @@ def _vgg_extractor(backbone_name: str, highres: bool, progress: bool, pretrained
 
     # find the index of the layer from which we wont freeze
     assert 0 <= trainable_layers <= num_stages
-    freeze_before = num_stages if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
+    freeze_before = len(backbone) if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
 
     for b in backbone[:freeze_before]:
         for parameter in b.parameters():
diff --git a/torchvision/models/detection/ssdlite.py b/torchvision/models/detection/ssdlite.py
index 30e4383ed5d..26378d7038d 100644
--- a/torchvision/models/detection/ssdlite.py
+++ b/torchvision/models/detection/ssdlite.py
@@ -147,7 +147,7 @@ def _mobilenet_extractor(backbone_name: str, progress: bool, pretrained: bool, t
 
     # find the index of the layer from which we wont freeze
     assert 0 <= trainable_layers <= num_stages
-    freeze_before = num_stages if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
+    freeze_before = len(backbone) if trainable_layers == 0 else stage_indices[num_stages - trainable_layers]
 
     for b in backbone[:freeze_before]:
         for parameter in b.parameters():