compatibility for torchvision 0.4.0

fcos_core/data/datasets/coco.py

@@ -61,7 +61,7 @@ class COCODataset(torchvision.datasets.coco.CocoDetection):
             v: k for k, v in self.json_category_id_to_contiguous_id.items()
         }
         self.id_to_img_map = {k: v for k, v in enumerate(self.ids)}
-        self.transforms = transforms
+        self._transforms = transforms
 
     def __getitem__(self, idx):
         img, anno = super(COCODataset, self).__getitem__(idx)
@@ -90,8 +90,8 @@ class COCODataset(torchvision.datasets.coco.CocoDetection):
 
         target = target.clip_to_image(remove_empty=True)
 
-        if self.transforms is not None:
-            img, target = self.transforms(img, target)
+        if self._transforms is not None:
+            img, target = self._transforms(img, target)
 
         return img, target, idx
 

fcos_core/data/transforms/transforms.py

@@ -57,9 +57,12 @@ class Resize(object):
     def __call__(self, image, target=None):
         size = self.get_size(image.size)
         image = F.resize(image, size)
-        if target is None:
+        if isinstance(target, list):
+            target = [t.resize(image.size) for t in target]
+        elif target is None:
             return image
-        target = target.resize(image.size)
+        else:
+            target = target.resize(image.size)
         return image, target
 
 

fcos_core/structures/segmentation_mask.py

@@ -414,7 +414,8 @@ class PolygonList(object):
         else:
             # advanced indexing on a single dimension
             selected_polygons = []
-            if isinstance(item, torch.Tensor) and item.dtype == torch.uint8:
+            if isinstance(item, torch.Tensor) and \
+               item.dtype == torch.uint8 or item.dtype == torch.bool:
                 item = item.nonzero()
                 item = item.squeeze(1) if item.numel() > 0 else item
                 item = item.tolist()