From ee87438c9395355b51946c225d01713dc32121ab Mon Sep 17 00:00:00 2001
From: fhasibi <fateme.hasibi.cbuk@gmail.com>
Date: Tue, 9 May 2023 14:36:02 +0330
Subject: [PATCH] fix #10 add train class

---
 model/easyOCR.py | 128 ++++++++++++++++++++---------------------------
 1 file changed, 55 insertions(+), 73 deletions(-)

diff --git a/model/easyOCR.py b/model/easyOCR.py
index 96f5680..985a437 100644
--- a/model/easyOCR.py
+++ b/model/easyOCR.py
@@ -2,82 +2,64 @@
 from easyocr import Reader
 from sklearn.model_selection import train_test_split
 import json
-from PIL import Image, ImageDraw , ImageFont
-
-
-
-# Direction of images and nnotation
-image_dir = "./business_card_dataset/images"
-annotation_path ="./business_card_dataset/annotations/instances_default.json"
-
-# Open and read the annotaion
-with open(annotation_path , 'r' , encoding='utf-8') as f:
-    annotations = json.load(f)
-
-# Set languages that gonna use in the project
-languages = ['en' , 'ja']
-
-# Split dataset into training and testing sets
-image_files = os.listdir(image_dir)
-train_files, test_files = train_test_split(image_files, test_size=0.1, random_state=42)
-
-# Create easyOCR reader classs
-reader = Reader(languages)
-
-# Loop thtough each image file
-for img in train_files:
-
-    # Check the extension(.PNG)
-    if img.endswith('.png'):
-
-        # Get the full path to the image and annotation file
-        image_path = os.path.join(image_dir , img)
-
-        image = os.path.basename(image_path)
-
-        # Extract text and bounding box from annotation
+from PIL import Image
+
+class easyOCR:
+    def __init__(self , languages = ['en' , 'ja'] , image_dir = "./business_card_dataset/images" , annotation_path ="./business_card_dataset/annotations/instances_default.json"):
+          self.languages = languages
+          self.image_dir = image_dir
+          self.annotation_path = annotation_path
+          self.reader  = Reader(languages)
+
+    # Open and read the annotaion
+    def get_annotations_and_image_file(self):
+        with open(self.annotation_path , 'r' , encoding='utf-8') as f:
+            annotations = json.load(f)
+        image_files = os.listdir(self.image_dir)
+        return annotations , image_files
+    
+    def extract_text_box(self , image_path , annotations):
         texts = []
         boxes = []
         for annotation in annotations["annotations"]:
-                if annotation.get('filename') == image and annotation["language"] in languages:
+                if annotation.get('filename') == os.path.basename(image_path) and annotation["language"] in self.languages:
                     boxes.append(annotation["bbox"])
                     texts.append(annotation["text"])
-
-        # Create a dictionary with the bounding boxes
-        detail = {'box': boxes}
-
-        # Train the model using image and annotation
-        results = reader.readtext(image_path , detail=detail)
+                return texts , boxes
         
-
-
-
-# Test the model using the testing set
-for img in test_files:
-    if img.endswith('.png'):
-        image_path = os.path.join(image_dir , img)
-
-        image = os.path.basename(image_path)
-        texts = []
-        boxes = []
-        for annotation in annotations["annotations"]:
-                if annotation.get('filename') == image and annotation["language"] in languages:
-                    boxes.append(annotation["bbox"])
-                    texts.append(annotation["text"])
-        detail = {'box': boxes}
-        results = reader.readtext(image_path , detail=detail)
-        for res in results:
-             text = res[1]
-             bbox = res[0]
-             conf = res[2]
-             print(f"System output : {text} - Confidence: {conf}")
-
-             # Crop text region
-             if all(isinstance(coord, int) for coord in bbox):
-                 bbox = tuple(int(coord) for coord in bbox)
-                 try:
-                     region = Image.open(image_path).crop(bbox)
-                     region.save(f"{text}.png")
-                 except ValueError:
-                     pass
-         
+    def train(self):
+        annotations , image_files = self.get_annotations_and_image_file()
+        train_files , test_files = train_test_split(image_files, test_size=0.1 , random_state=42)
+        for img in train_files:
+            if img.endswith('.png'):
+                image_path = os.path.join(self.image_dir , img)
+                texts, boxes = self.extract_text_box(image_path, annotations)
+                detail = {'box': boxes}
+                results = self.reader.readtext(image_path , detail=detail)
+
+    def test(self):
+        annotations , image_files = self.get_annotations_and_image_file()
+        train_files , test_files = train_test_split(image_files, test_size=0.1 , random_state=42)
+        for img in test_files:
+            if img.endswith('.png'):
+                image_path = os.path.join(self.image_dir , img)
+                texts, boxes = self.extract_text_box(image_path, annotations)
+                detail = {'box': boxes}
+                results = self.reader.readtext(image_path , detail=detail)
+                for res in results:
+                    text = res[1]
+                    bbox = res[0]
+                    conf = res[2]
+                    print(f"System output : {text} - Confidence: {conf}")
+                    if all(isinstance(coord, int) for coord in bbox):
+                        bbox = tuple(int(coord) for coord in bbox)
+                        try:
+
+                            region = Image.open(image_path).crop(bbox)
+                            region.save(f"{text}.png")
+                        except ValueError:
+                           pass
+
+model = easyOCR()
+model.train()
+model.test()