adding argument `value_type` = 'float_list' forces it to convert any `int`...

adding argument `value_type` = 'float_list' forces it to convert any `int` values to float. Tensorflow ops requires this argument to be a float whereas some annotations tools are wrongly mentioning `area` as `int64` dtype.

PiperOrigin-RevId: 475385551

official/vision/data/create_coco_tf_record.py

@@ -226,7 +226,7 @@ def bbox_annotations_to_feature_dict(
       'image/object/is_crowd':
           tfrecord_lib.convert_to_feature(data['is_crowd']),
       'image/object/area':
-          tfrecord_lib.convert_to_feature(data['area']),
+          tfrecord_lib.convert_to_feature(data['area'], 'float_list')
   }
   if include_masks:
     feature_dict['image/object/mask'] = (