Added exception tests for CustomCropResize

Signed-off-by: Martin <[email protected]>

scripts/train_torchreid.py

@@ -38,7 +38,7 @@
     # Specify ReID model
     model = build_model(
         name="osnet_x1_0",
-        num_classes=6878,  # max person id of PT21
+        num_classes=6879,  # max person id of PT21
         # use_gpu=True,
         # pretrained=True,
         # loss="softmax",

tests/utils/test__image.py

@@ -22,7 +22,7 @@
 )
 from dgs.utils.types import ImgShape, TVImage
 from dgs.utils.validation import validate_bboxes, validate_images, validate_key_points
-from helper import load_test_image
+from helper import load_test_image, load_test_images
 
 # Map image name to shape.
 # Shape is torch shape and therefore [C x H x W].
@@ -654,6 +654,23 @@ def test_batched_input(self):
                         self.assertTrue(torch.allclose(new_bboxes[0], new_bboxes[1]))
                         self.assertTrue(torch.allclose(new_coords[0], new_coords[1]))
 
+    def test_exceptions(self):
+        for images, bboxes, coords, exception in [
+            (load_test_image("866-200x300.jpg"), create_bbox(10, 10), torch.zeros((10, 21, 2)), ValueError),
+            (
+                load_test_images(["866-200x300.jpg", "866-200x300.jpg"]),
+                tv_tensors.BoundingBoxes(torch.zeros((10, 4)), canvas_size=(10, 10), format="xywh"),
+                torch.zeros((10, 21, 2)),
+                ValueError,
+            ),
+        ]:
+            with self.subTest(msg=f"bboxes, coords, exception"):
+                data = create_structured_data(
+                    image=images, out_shape=(100, 100), mode="zero-pad", bbox=bboxes, key_points=coords
+                )
+                with self.assertRaises(exception):
+                    CustomCropResize()(data)
+
 
 if __name__ == "__main__":
     unittest.main()