[datasets] read labels in utf-8 and log input string on vocab error (m…

…indee#1479)

doctr/datasets/recognition.py

@@ -37,7 +37,7 @@ def __init__(
         super().__init__(img_folder, **kwargs)
 
         self.data: List[Tuple[str, str]] = []
-        with open(labels_path) as f:
+        with open(labels_path, encoding="utf-8") as f:
             labels = json.load(f)
 
         for img_name, label in labels.items():

doctr/datasets/utils.py

@@ -78,7 +78,10 @@ def encode_string(
     try:
         return list(map(vocab.index, input_string))
     except ValueError:
-        raise ValueError("some characters cannot be found in 'vocab'")
+        raise ValueError(
+            f"some characters cannot be found in 'vocab'. \
+                         Please check the input string {input_string} and the vocabulary {vocab}"
+        )
 
 
 def decode_sequence(

references/recognition/README.md

@@ -75,7 +75,7 @@ The order of entries in the json does not matter.
 }
 ```
 
-When typing your labels, be aware that the VOCAB doesn't handle spaces.
+When typing your labels, be aware that the VOCAB doesn't handle spaces. Also make sure your `labels.json` file is using UTF-8 encoding.
 
 ## Advanced options