diff --git a/recognizer/crnn/lib/create_lmdb_dataset.py b/recognizer/crnn/lib/create_lmdb_dataset.py
index 79dde34..f532126 100644
--- a/recognizer/crnn/lib/create_lmdb_dataset.py
+++ b/recognizer/crnn/lib/create_lmdb_dataset.py
@@ -1,4 +1,3 @@
-
 import lmdb
 import cv2
 import numpy as np
@@ -26,7 +25,9 @@ def checkImageIsValid(imageBin):
 
 def writeCache(env, cache):
     with env.begin(write=True) as txn:
-        for k, v in cache.items():
+        for _k, _v in cache.items():
+            k = _k.encode() if type(_k) == str else _k
+            v = _v.encode() if type(_v) == str else _v
             txn.put(k, v)
 
 
@@ -54,7 +55,7 @@ def createDataset(outputPath, imagePathList, labelList, lexiconList=None, checkV
         #     print('%s does not exist' % imagePath)
         #     continue
 
-        with open(imagePath, 'r') as f:
+        with open(imagePath, 'rb') as f:
             imageBin = f.read()
 
         if checkValid:
diff --git a/recognizer/crnn/lib/dataset.py b/recognizer/crnn/lib/dataset.py
index 436a918..fc8d63a 100644
--- a/recognizer/crnn/lib/dataset.py
+++ b/recognizer/crnn/lib/dataset.py
@@ -60,7 +60,7 @@ def __getitem__(self, index):
                 img = self.transform(img)
 
             label_key = 'label-%09d' % index
-            label = txn.get(label_key.encode())
+            label = txn.get(label_key.encode()).decode()
 
             if self.target_transform is not None:
                 label = self.target_transform(label)
diff --git a/recognizer/crnn/lib/get_alphabets.py b/recognizer/crnn/lib/get_alphabets.py
new file mode 100644
index 0000000..f17fb70
--- /dev/null
+++ b/recognizer/crnn/lib/get_alphabets.py
@@ -0,0 +1,22 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+__title__ = ''
+__author__ = 'changxin'
+__mtime__ = '2019/3/14'
+"""
+
+
+def get_alphabets(filelists):
+    a = []
+    for x in filelists:
+        with open(x) as f:
+            strings = f.readlines()
+            string = [y.strip().split(' ')[1] for y in strings]
+            tmp = []
+            for z in ''.join(string):
+                tmp.append(z)
+            a = a + tmp
+    with open('alphabets.py', 'w') as e:
+        e.write("#coding=utf8\n")
+        e.write('\"\"\"' + ''.join(list(set(a))) + '\"\"\"')