Surya Recognizer

[titipata]

ตัวอย่างการใช้ผลจาก YOLO dectector ดึงส่วนของภาพออกมาและใช้ surya-ocr ในการอ่านข้อความ

from pathlib import Path
from ultralytics import YOLO

checkpoint = Path("best_detection_yolo.pt").resolve()
# Load a model from a checkpoint
model = YOLO(checkpoint)
results = model.predict(images, imgsz=640, conf=0.25, half=True)

from surya.recognition import batch_recognition
from surya.model.recognition.model import load_model as load_recognizer
from surya.model.recognition.processor import load_processor as load_recognizer_processor

recognizer = load_recognizer()
recognizer_processor = load_recognizer_processor()

# Process each YOLO result
padding = 0
images_for_ocr = []
languages_for_ocr = []
for idx, result in enumerate(results):
    orig_img = result.orig_img
    if result.boxes is not None:
        boxes = result.boxes.xyxy.cpu().numpy()
        confidences = result.boxes.conf.cpu().numpy()
        for box, conf in zip(boxes, confidences):
            # Extract coordinates with padding
            x1, y1, x2, y2 = map(int, box)
            x1 = max(0, x1 - padding)
            y1 = max(0, y1 - padding)
            x2 = min(orig_img.shape[1], x2 + padding)
            y2 = min(orig_img.shape[0], y2 + padding)
            region = orig_img[y1:y2, x1:x2]
            pil_region = Image.fromarray(region)
            images_for_ocr.append(pil_region)
languages_for_ocr = [["th", "en"]] * len(images_for_ocr)

text_results, confidence_scores = batch_recognition(
    images=images_for_ocr,
    languages=languages_for_ocr,
    model=recognizer,
    processor=recognizer_processor
)

[titipata]

Matching (alternative)

from fuzzywuzzy import process
process.extractOne(
    "ดีเสล", ["ดีเซล", "เบนซิน"], score_cutoff = 60
)