diff --git a/deeplabcut/pose_estimation_pytorch/data/dataset.py b/deeplabcut/pose_estimation_pytorch/data/dataset.py
index cbb9b9726a..c2e0da9891 100644
--- a/deeplabcut/pose_estimation_pytorch/data/dataset.py
+++ b/deeplabcut/pose_estimation_pytorch/data/dataset.py
@@ -250,7 +250,7 @@ def _prepare_final_annotation_dict(
 
         # we use ..., :3 to pass the visibility flag along
         return {
-            "keypoints": pad_to_length(keypoints[..., :3], num_animals, -1).astype(
+            "keypoints": pad_to_length(keypoints[..., :3], num_animals, 0).astype(
                 np.single
             ),
             "keypoints_unique": keypoints_unique[..., :3].astype(np.single),
diff --git a/deeplabcut/pose_estimation_pytorch/data/transforms.py b/deeplabcut/pose_estimation_pytorch/data/transforms.py
index 6e5bd0d1e9..6d14870bd7 100644
--- a/deeplabcut/pose_estimation_pytorch/data/transforms.py
+++ b/deeplabcut/pose_estimation_pytorch/data/transforms.py
@@ -78,7 +78,7 @@ def build_transforms(augmentations: dict) -> A.BaseCompose:
         if rotation is not None:
             rotation = (-rotation, rotation)
         if translation is not None:
-            translation = (0, translation)
+            translation = (-translation, translation)
 
         transforms.append(
             A.Affine(
@@ -281,6 +281,8 @@ def get_params_dependent_on_targets(self, params: dict[str, Any]) -> dict[str, A
         sampling = self.crop_sampling
         if self.crop_sampling == "hybrid":
             sampling = np.random.choice(["uniform", "density"])
+        if len(kpts) == 0:
+            sampling = "uniform"
         if sampling == "uniform":
             center = np.random.random(2)
         else:
@@ -310,7 +312,7 @@ def apply_to_keypoints(
         self,
         keypoints,
         **params,
-    ) -> list[float]:
+    ) -> list[tuple[float]]:
         keypoints = super().apply_to_keypoints(keypoints, **params)
         new_keypoints = []
         for kp in keypoints: