app.py

import os
import sys
import json
import numpy as np
import cv2
import traceback
import shutil
import gradio as gr

os.makedirs('models/textdetector', exist_ok=True)
os.makedirs('models/ocr', exist_ok=True)
os.makedirs('models/inpainting', exist_ok=True)

models_mapping = {
    'comictextdetector.pt': 'models/textdetector/comictextdetector.pt',
    'mit48pxctc_ocr.ckpt': 'models/ocr/mit48px_ctc.ckpt',
    'lama_large_512px.ckpt': 'models/inpainting/lama_large_512px.ckpt'
}

for src, dst in models_mapping.items():
    if os.path.exists(src) and not os.path.exists(dst):
        try:
            shutil.move(src, dst)
        except Exception:
            pass

sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))

detector_model = None
ocr_model = None

# ==========================================
# دالة التحديد (ترجع مضلعات)
# ==========================================
def detect_boxes(image):
    global detector_model
    try:
        if detector_model is None:
            from modules.base import load_modules
            load_modules()
            from modules.textdetector import TEXTDETECTORS
            
            if 'ctd' in TEXTDETECTORS.module_dict:
                detector_model = TEXTDETECTORS.module_dict['ctd']()
            else:
                return json.dumps([{"error": "موديل التحديد ctd لم يتم تحميله. تأكد من إضافة setuptools في requirements"}])
            
            if hasattr(detector_model, 'load_model') and not detector_model.all_model_loaded():
                detector_model.load_model()
        
        if image.shape[2] == 3:
            image_bgr = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
        else:
            image_bgr = image

        mask, boxes = detector_model.detect(image_bgr)
        real_boxes = []
        
        for i, b in enumerate(boxes):
            pts = None
            if hasattr(b, 'box') and len(b.box) == 4:
                x, y = b.box[0], b.box[1]
                w = b.box[2] - b.box[0]
                h = b.box[3] - b.box[1]
                pts = np.array([[x, y], [x + w, y], [x + w, y + h], [x, y + h]], dtype=np.float32)
            elif hasattr(b, 'pts') or hasattr(b, 'lines'):
                pts = np.array(b.pts if hasattr(b, 'pts') else b.lines, dtype=np.float32)
                
            if pts is None: continue
            
            x, y, w, h = cv2.boundingRect(pts)
            real_boxes.append({
                "id": i, 
                "x": int(x), "y": int(y), "w": int(w), "h": int(h),
                "polygon": pts.tolist() # إرسال المضلع لصفحة الويب
            })
            
        return json.dumps(real_boxes)
    except Exception as e:
        return json.dumps([{"error": f"===== خطأ في التحديد =====\n\n{traceback.format_exc()}"}])

# ==========================================
# دالة الاستخراج (تستقبل مضلعات)
# ==========================================
def extract_text(image, boxes_json):
    global ocr_model
    try:
        if not boxes_json or str(boxes_json).strip() == "":
            return "لم يتم إرسال أي مربعات! ارسم أو حدد المربعات أولاً."

        boxes_data = json.loads(boxes_json)
        if not boxes_data: return "لم يتم العثور على مربعات."
        
        if len(boxes_data) > 0 and 'error' in boxes_data[0]:
            return boxes_data[0]['error']

        from utils.textblock import TextBlock
        blk_list = []
        for b in boxes_data:
            if 'error' in b: continue
            blk = TextBlock()
            
            # قراءة المضلع من صفحة الويب
            if 'polygon' in b and b['polygon']:
                pts = np.array(b['polygon'], dtype=np.float32)
                x, y, w, h = cv2.boundingRect(pts)
                blk.box = [x, y, x + w, y + h]
                blk.polygon = pts
                blk.lines = [pts]
            else:
                x, y, w, h = int(b['x']), int(b['y']), int(b['w']), int(b['h'])
                blk.box = [x, y, x + w, y + h]
                pts = np.array([[x, y], [x + w, y], [x + w, y + h], [x, y + h]], dtype=np.float32)
                blk.polygon = pts
                blk.lines = [pts] 
            
            blk_list.append(blk)
        
        if ocr_model is None:
            from modules.base import load_modules
            load_modules()
            from modules.ocr import OCR
            
            target_key = None
            for key in OCR.module_dict.keys():
                if 'ctc' in key.lower() or '48' in key.lower():
                    target_key = key
                    break
            
            if not target_key and OCR.module_dict:
                target_key = list(OCR.module_dict.keys())[0]

            ocr_model = OCR.module_dict[target_key]()
            
            if hasattr(ocr_model, 'all_model_loaded') and not ocr_model.all_model_loaded():
                ocr_model.load_model()

        if image.shape[2] == 3:
            image_bgr = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
        else:
            image_bgr = image

        results = ocr_model.run_ocr(image_bgr, blk_list)
        
        final_text = []
        if results:
            for r in results:
                if isinstance(r.text, list): final_text.append(" ".join(r.text))
                else: final_text.append(str(r.text))
        
        res_text = "\n\n".join(final_text)
        
        if not res_text.strip():
            return "الموديل لم يستطع قراءة أي نص. (تأكد أن المضلع محيط بالنص بشكل صحيح)"
            
        return res_text
        
    except Exception as e:
        return f"===== خطأ برمجي =====\n\n{traceback.format_exc()}"

with gr.Blocks() as demo:
    img_input = gr.Image(type="numpy")
    boxes_input = gr.Textbox()
    detect_out = gr.Textbox()
    extract_out = gr.Textbox()
    
    detect_btn = gr.Button("detect")
    extract_btn = gr.Button("extract")
    
    detect_btn.click(fn=detect_boxes, inputs=img_input, outputs=detect_out, api_name="detect")
    extract_btn.click(fn=extract_text, inputs=[img_input, boxes_input], outputs=extract_out, api_name="extract")

if __name__ == "__main__":
    demo.launch()