fix(types): fix location and bounding box in converters (#3847)

* fix(types): fix location and bounding box in converters

* fix(types): fix location and bounding box in converters

jina/types/document/converters.py

@@ -544,14 +544,17 @@ def convert_image_blob_to_sliding_windows(
             strides=(row_step * stride_h, col_step * stride_w, row_step, col_step, 1),
             writeable=False,
         )
-
+        cur_loc_h, cur_loc_w = 0, 0
+        if self.location:
+            cur_loc_h, cur_loc_w = self.location[:2]
+
+        bbox_locations = [
+            (h * stride_h + cur_loc_h, w * stride_w + cur_loc_w, window_h, window_w)
+            for h in range(expanded_img.shape[0])
+            for w in range(expanded_img.shape[1])
+        ]
         expanded_img = expanded_img.reshape((-1, window_h, window_w, c))
         if as_chunks:
-            bbox_locations = [
-                (h * stride_h, w * stride_w)
-                for h in range(expanded_img.shape[0])
-                for w in range(expanded_img.shape[1])
-            ]
             from . import Document
 
             for location, _blob in zip(bbox_locations, expanded_img):