Minor fixes

.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+*.pyc

utils.py

@@ -1,7 +1,7 @@
 from doctr.models import detection_predictor, recognition_predictor
 from doctr.io import DocumentFile
-from surya.recognition import RecognitionPredictor
-from surya.detection import DetectionPredictor
+from surya.recognition import RecognitionPredictor as SuryaRecognitionPredictor
+from surya.detection import DetectionPredictor as SuryaDetectionPredictor
 from PIL import Image
 # from functools import lru_cache
 from torchvision import models
@@ -246,17 +246,17 @@ def predict_ne(image_path, device="cpu"):
         return le.inverse_transform([predicted.item()])[0]
 
 doctr_detector = None
-recognition_predictor = None
-detection_predictor = None
+surya_recognition_predictor = None
+surya_detection_predictor = None
 def initialize_detector():
-    global doctr_detector, recognition_predictor, detection_predictor
+    global doctr_detector, surya_recognition_predictor, surya_detection_predictor
     if doctr_detector is None:
         doctr_detector = detection_predictor('db_mobilenet_v3_large', pretrained=True, assume_straight_pages=True, preserve_aspect_ratio=True)
-    if recognition_predictor is None:
-        recognition_predictor = RecognitionPredictor()
-    if detection_predictor is None:
-        detection_predictor = DetectionPredictor()
-    return doctr_detector, recognition_predictor, detection_predictor
+    if surya_recognition_predictor is None:
+        surya_recognition_predictor = SuryaRecognitionPredictor()
+    if surya_detection_predictor is None:
+        surya_detection_predictor = SuryaDetectionPredictor()
+    return doctr_detector, surya_recognition_predictor, surya_detection_predictor
 
 def get_cleaned_boxes(out, page):
     h, w, _ = page.shape
@@ -324,7 +324,7 @@ def merge_boxes_same_line(boxes, y_thresh=5, x_thresh=60):
     return np.array(merged)
 
 def ocr_citizenship_utils(image_path: str) -> List[List[str]]:
-    doctr_detector, recognition_predictor, detection_predictor = initialize_detector()
+    doctr_detector, surya_recognition_predictor, surya_detection_predictor = initialize_detector()
     page = cv2.imread(image_path)
     page = cv2.convertScaleAbs(page, alpha=1.5, beta=0)
     page = cv2.resize(page, (720,480))
@@ -347,7 +347,7 @@ def ocr_citizenship_utils(image_path: str) -> List[List[str]]:
 
         # OCR PART
         langs = ["en",'ne']
-        predictions = recognition_predictor_surya([pil_image], [langs], detection_predictor_surya)
+        predictions = surya_recognition_predictor([pil_image], [langs])
         text_combo = ''
         for text_line in predictions[0].text_lines:
             text_combo = text_combo + " " + text_line.text.strip()