fix synthtext dataset convert (#807)

mindspore-lab · Feb 14, 2025 · 55e0ee7 · 55e0ee7
1 parent fecd8ed
commit 55e0ee7
Showing 1 changed file with 3 additions and 3 deletions.
diff --git a/tools/dataset_converters/synthtext.py b/tools/dataset_converters/synthtext.py
@@ -143,7 +143,7 @@ def convert_rec_lmdb(self, image_dir=None, label_path=None, output_path=None):
         wordBB, txt = zip(*data_list)
         for i in range(len(mat["wordBB"][0])):
             mat["wordBB"][0][i] = wordBB[i]
-        mat["txt"] = np.array(txt).reshape(1, -1)
+        mat["txt"] = np.array(txt, dtype=object).reshape(1, -1)
 
         data_list = defaultdict(list)
         for image, polys, texts in zip(mat["imnames"][0], mat["wordBB"][0], mat["txt"][0]):
@@ -168,8 +168,8 @@ def convert_rec_lmdb(self, image_dir=None, label_path=None, output_path=None):
             )
 
         images, labels = zip(*data_list)
-        images = iter(itertools.chain(*images))
-        labels = iter(itertools.chain(*labels))
+        images = list(itertools.chain(*images))
+        labels = list(itertools.chain(*labels))
 
         print("Creating the LMDB dataset.")
         create_lmdb_dataset(images, labels, output_path=output_path)