diff --git a/easyocr/easyocr.py b/easyocr/easyocr.py
index 681c05b3c..0fab10842 100644
--- a/easyocr/easyocr.py
+++ b/easyocr/easyocr.py
@@ -1,6 +1,6 @@
 # -*- coding: utf-8 -*-
 
-from .recognition import get_recognizer, get_text
+from .recognition import get_recognizer, get_text, get_recognizer_attn
 from .utils import group_text_box, get_image_list, calculate_md5, get_paragraph,\
                    download_and_unzip, printProgressBar, diff, reformat_input,\
                    make_rotated_img_list, set_result_with_confidence,\
@@ -228,9 +228,15 @@ def __init__(self, lang_list, gpu=True, model_storage_directory=None,
                     }
             else:
                 network_params = recog_config['network_params']
-            self.recognizer, self.converter = get_recognizer(recog_network, network_params,\
+            if "Prediction" in recog_config:
+                if recog_config["Prediction"] == "Attn":
+                    self.recognizer, self.converter = get_recognizer_attn(recog_network, network_params,\
                                                          self.character, separator_list,\
                                                          dict_list, model_path, device = self.device, quantize=quantize)
+            else:
+                self.recognizer, self.converter = get_recognizer(recog_network, network_params,\
+                                                            self.character, separator_list,\
+                                                            dict_list, model_path, device = self.device, quantize=quantize)
 
     def getDetectorPath(self, detect_network):
         if detect_network in self.support_detection_network:
diff --git a/easyocr/recognition.py b/easyocr/recognition.py
index 147370f7d..f10850c28 100644
--- a/easyocr/recognition.py
+++ b/easyocr/recognition.py
@@ -8,6 +8,7 @@
 from collections import OrderedDict
 import importlib
 from .utils import CTCLabelConverter
+from ..trainer.utils import AttnLabelConverter
 import math
 
 def custom_mean(x):
@@ -121,21 +122,26 @@ def recognizer_predict(model, converter, test_loader, batch_max_length,\
             preds_prob = preds_prob/np.expand_dims(pred_norm, axis=-1)
             preds_prob = torch.from_numpy(preds_prob).float().to(device)
 
-            if decoder == 'greedy':
-                # Select max probabilty (greedy decoding) then decode index to character
+            if isinstance(converter, AttnLabelConverter):
                 _, preds_index = preds_prob.max(2)
-                preds_index = preds_index.view(-1)
-                preds_str = converter.decode_greedy(preds_index.data.cpu().detach().numpy(), preds_size.data)
-            elif decoder == 'beamsearch':
-                k = preds_prob.cpu().detach().numpy()
-                preds_str = converter.decode_beamsearch(k, beamWidth=beamWidth)
-            elif decoder == 'wordbeamsearch':
-                k = preds_prob.cpu().detach().numpy()
-                preds_str = converter.decode_wordbeamsearch(k, beamWidth=beamWidth)
+                preds_str = converter.decode(preds_index, preds_size.data)
+            else:
+                if decoder == 'greedy':
+                    # Select max probabilty (greedy decoding) then decode index to character
+                    _, preds_index = preds_prob.max(2)
+                    preds_index = preds_index.view(-1)
+                    preds_str = converter.decode_greedy(preds_index.data.cpu().detach().numpy(), preds_size.data)
+                elif decoder == 'beamsearch':
+                    k = preds_prob.cpu().detach().numpy()
+                    preds_str = converter.decode_beamsearch(k, beamWidth=beamWidth)
+                elif decoder == 'wordbeamsearch':
+                    k = preds_prob.cpu().detach().numpy()
+                    preds_str = converter.decode_wordbeamsearch(k, beamWidth=beamWidth)
 
             preds_prob = preds_prob.cpu().detach().numpy()
             values = preds_prob.max(axis=2)
             indices = preds_prob.argmax(axis=2)
+
             preds_max_prob = []
             for v,i in zip(values, indices):
                 max_probs = v[i!=0]
@@ -145,7 +151,15 @@ def recognizer_predict(model, converter, test_loader, batch_max_length,\
                     preds_max_prob.append(np.array([0]))
 
             for pred, pred_max_prob in zip(preds_str, preds_max_prob):
-                confidence_score = custom_mean(pred_max_prob)
+                
+
+                if isinstance(converter, AttnLabelConverter):
+                    pred_EOS = pred.find('[s]')
+                    pred = pred[:pred_EOS]
+                    confidence_score = custom_mean(pred_max_prob[:pred_EOS])
+                else:
+                    confidence_score = custom_mean(pred_max_prob)
+
                 result.append([pred, confidence_score])
 
     return result
@@ -153,7 +167,6 @@ def recognizer_predict(model, converter, test_loader, batch_max_length,\
 def get_recognizer(recog_network, network_params, character,\
                    separator_list, dict_list, model_path,\
                    device = 'cpu', quantize = True):
-
     converter = CTCLabelConverter(character, separator_list, dict_list)
     num_class = len(converter.character)
 
@@ -183,6 +196,38 @@ def get_recognizer(recog_network, network_params, character,\
 
     return model, converter
 
+def get_recognizer_attn(recog_network, network_params, character,\
+                   separator_list, dict_list, model_path,\
+                   device = 'cpu', quantize = True):
+    converter = AttnLabelConverter(character, device)
+    num_class = len(converter.character)
+
+    if recog_network == 'generation1':
+        model_pkg = importlib.import_module("easyocr.model.model")
+    elif recog_network == 'generation2':
+        model_pkg = importlib.import_module("easyocr.model.vgg_model")
+    else:
+        model_pkg = importlib.import_module(recog_network)
+    model = model_pkg.Model(num_class=num_class, **network_params)
+
+    if device == 'cpu':
+        state_dict = torch.load(model_path, map_location=device)
+        new_state_dict = OrderedDict()
+        for key, value in state_dict.items():
+            new_key = key[7:]
+            new_state_dict[new_key] = value
+        model.load_state_dict(new_state_dict)
+        if quantize:
+            try:
+                torch.quantization.quantize_dynamic(model, dtype=torch.qint8, inplace=True)
+            except:
+                pass
+    else:
+        model = torch.nn.DataParallel(model).to(device)
+        model.load_state_dict(torch.load(model_path, map_location=device))
+
+    return model, converter
+
 def get_text(character, imgH, imgW, recognizer, converter, image_list,\
              ignore_char = '',decoder = 'greedy', beamWidth =5, batch_size=1, contrast_ths=0.1,\
              adjust_contrast=0.5, filter_ths = 0.003, workers = 1, device = 'cpu'):
diff --git a/recognize_function.py b/recognize_function.py
new file mode 100644
index 000000000..819bb1a6f
--- /dev/null
+++ b/recognize_function.py
@@ -0,0 +1,41 @@
+import easyocr
+import os
+import cv2
+import numpy as np
+from easyocr.easyocr import Reader
+from easyocr.trainer.utils import AttnLabelConverter
+
+def recognize_text_from_images(image_pieces, models_directory, recog_network='best_accuracy', gpu=False):
+    """
+    Recognizes text from a list of image pieces using EasyOCR.
+
+    Parameters:
+    - image_pieces (list): List of image pieces as PIL Image objects.
+    - models_directory (str): Path to the models directory.
+    - recog_network (str): Recognition network to use (default is 'best_accuracy').
+    - gpu (bool): Whether to use GPU for OCR (default is False).
+
+    Returns:
+    - List of recognized texts.
+    """
+    model_storage_directory = os.path.join(models_directory, "model")
+    user_network_directory = os.path.join(models_directory, "user_network")
+
+    # Initialize EasyOCR reader
+    reader = Reader(['ru'], recog_network=recog_network, gpu=gpu,
+                            model_storage_directory=model_storage_directory,
+                            user_network_directory=user_network_directory)
+    
+
+    recognized_texts = []
+    for image_piece in image_pieces:
+        # Convert PIL Image to OpenCV format
+        image_cv = cv2.cvtColor(np.array(image_piece), cv2.COLOR_RGB2BGR)
+        # Perform text recognition
+        if isinstance(reader.converter, AttnLabelConverter):
+            result = reader.readtext(image_cv, detail=0, decoder="beamsearch")
+        else:
+            result = reader.readtext(image_cv, detail=0)
+        recognized_texts.append(" ".join(result))
+    
+    return recognized_texts
\ No newline at end of file
diff --git a/test.py b/test.py
new file mode 100644
index 000000000..5a00586f0
--- /dev/null
+++ b/test.py
@@ -0,0 +1,59 @@
+import easyocr
+import matplotlib.pyplot as plt
+import cv2
+from PIL import Image, ImageDraw, ImageFont
+import numpy as np
+import os
+
+
+# Load the image
+image_path = "test2 (1).jpg"
+image = cv2.imread(image_path)
+image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+
+# Convert to PIL Image for better font rendering
+image_pil = Image.fromarray(image_rgb)
+draw = ImageDraw.Draw(image_pil)
+
+# Load a font (make sure the path to the font file is correct)
+font_path = "arial.ttf"  # Replace with the path to your .ttf font file
+font = ImageFont.truetype(font_path, 60)  # Increased font size to 40
+
+models_directory = "C:/Users/user/Desktop/modelsocr"
+model_storage_directory = os.path.join(models_directory, "model")
+user_network_directory = os.path.join(models_directory, "user_network")
+# Initialize EasyOCR reader
+reader = easyocr.Reader(['ru'], recog_network='best_accuracy', gpu = False,
+                        model_storage_directory=model_storage_directory,
+                        user_network_directory=user_network_directory)
+
+# Perform text detection and recognition
+results = reader.readtext(image_path)
+
+# Print the results
+print(results)
+
+# Draw bounding boxes and placeholder text on the image
+for (bbox, text, prob) in results:
+    # Replace the recognized text with "вопросы"
+    
+    # Unpack the bounding box coordinates
+    (top_left, top_right, bottom_right, bottom_left) = bbox
+    top_left = tuple([int(val) for val in top_left])
+    bottom_right = tuple([int(val) for val in bottom_right])
+
+    # Draw the bounding box on the image
+    draw.rectangle([top_left, bottom_right], outline="green", width=2)
+
+    # Put the placeholder text
+    draw.text((top_left[0], top_left[1] - 40), text, font=font, fill="green")  # Adjusted y-coordinate for larger font
+
+# Convert back to OpenCV format
+image_rgb = cv2.cvtColor(np.array(image_pil), cv2.COLOR_RGB2BGR)
+
+# Display the image with bounding boxes and placeholder text
+plt.figure(figsize=(60, 60))  # Adjusted the figure size for better visibility
+plt.imshow(cv2.cvtColor(image_rgb, cv2.COLOR_BGR2RGB))
+plt.axis('off')
+plt.savefig("text.png")
+plt.show()
\ No newline at end of file
diff --git a/trainer/modules/prediction.py b/trainer/modules/prediction.py
index c8a40af0e..2b11f24e6 100644
--- a/trainer/modules/prediction.py
+++ b/trainer/modules/prediction.py
@@ -1,22 +1,23 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+#device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
 
 class Attention(nn.Module):
 
-    def __init__(self, input_size, hidden_size, num_classes):
+    def __init__(self, input_size, hidden_size, num_classes, device):
         super(Attention, self).__init__()
         self.attention_cell = AttentionCell(input_size, hidden_size, num_classes)
         self.hidden_size = hidden_size
         self.num_classes = num_classes
         self.generator = nn.Linear(hidden_size, num_classes)
+        self.device = device
 
     def _char_to_onehot(self, input_char, onehot_dim=38):
         input_char = input_char.unsqueeze(1)
         batch_size = input_char.size(0)
-        one_hot = torch.FloatTensor(batch_size, onehot_dim).zero_().to(device)
+        one_hot = torch.FloatTensor(batch_size, onehot_dim).zero_().to(self.device)
         one_hot = one_hot.scatter_(1, input_char, 1)
         return one_hot
 
@@ -30,9 +31,9 @@ def forward(self, batch_H, text, is_train=True, batch_max_length=25):
         batch_size = batch_H.size(0)
         num_steps = batch_max_length + 1  # +1 for [s] at end of sentence.
 
-        output_hiddens = torch.FloatTensor(batch_size, num_steps, self.hidden_size).fill_(0).to(device)
-        hidden = (torch.FloatTensor(batch_size, self.hidden_size).fill_(0).to(device),
-                  torch.FloatTensor(batch_size, self.hidden_size).fill_(0).to(device))
+        output_hiddens = torch.FloatTensor(batch_size, num_steps, self.hidden_size).fill_(0).to(self.device)
+        hidden = (torch.FloatTensor(batch_size, self.hidden_size).fill_(0).to(self.device),
+                  torch.FloatTensor(batch_size, self.hidden_size).fill_(0).to(self.device))
 
         if is_train:
             for i in range(num_steps):
@@ -44,8 +45,8 @@ def forward(self, batch_H, text, is_train=True, batch_max_length=25):
             probs = self.generator(output_hiddens)
 
         else:
-            targets = torch.LongTensor(batch_size).fill_(0).to(device)  # [GO] token
-            probs = torch.FloatTensor(batch_size, num_steps, self.num_classes).fill_(0).to(device)
+            targets = torch.LongTensor(batch_size).fill_(0).to(self.device)  # [GO] token
+            probs = torch.FloatTensor(batch_size, num_steps, self.num_classes).fill_(0).to(self.device)
 
             for i in range(num_steps):
                 char_onehots = self._char_to_onehot(targets, onehot_dim=self.num_classes)
diff --git a/trainer/utils.py b/trainer/utils.py
index 4cf7dac8c..75b021329 100644
--- a/trainer/utils.py
+++ b/trainer/utils.py
@@ -1,7 +1,7 @@
 import torch
 import pickle
 import numpy as np
-device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
 
 class AttrDict(dict):
     def __init__(self, *args, **kwargs):
@@ -298,12 +298,13 @@ def decode_wordbeamsearch(self, mat, beamWidth=5):
 class AttnLabelConverter(object):
     """ Convert between text-label and text-index """
 
-    def __init__(self, character):
+    def __init__(self, character, device):
         # character (str): set of the possible characters.
         # [GO] for the start token of the attention decoder. [s] for end-of-sentence token.
         list_token = ['[GO]', '[s]']  # ['[s]','[UNK]','[PAD]','[GO]']
         list_character = list(character)
         self.character = list_token + list_character
+        self.device = device
 
         self.dict = {}
         for i, char in enumerate(self.character):
@@ -331,7 +332,7 @@ def encode(self, text, batch_max_length=25):
             text.append('[s]')
             text = [self.dict[char] for char in text]
             batch_text[i][1:1 + len(text)] = torch.LongTensor(text)  # batch_text[:, 0] = [GO] token
-        return (batch_text.to(device), torch.IntTensor(length).to(device))
+        return (batch_text.to(self.device), torch.IntTensor(length).to(self.device))
 
     def decode(self, text_index, length):
         """ convert text-index into text-label. """