Update

layout_extractor.py

@@ -1,16 +1,16 @@
 import layoutparser as lp
 import pytesseract
-import json
 from pdf2image import convert_from_path
 from PIL import Image
 
 def convert_pdf_to_images(pdf_path):
     return convert_from_path(pdf_path)
 
+# ✅ Use EfficientDet instead of Detectron2 for better compatibility
 def analyze_layout(image):
-    model = lp.Detectron2LayoutModel(
-        config_path='lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config',
-        extra_config=["MODEL.ROI_HEADS.SCORE_THRESH_TEST", 0.8],
+    model = lp.EfficientDetLayoutModel(
+        "lp://efficientdet/PubLayNet",
+        extra_config=["MODEL.ROI_HEADS.SCORE_THRESH_TEST", 0.6],
         label_map={0: "Text", 1: "Title", 2: "List", 3: "Table", 4: "Figure"}
     )
     layout = model.detect(image)
@@ -31,11 +31,11 @@ def extract_text_from_blocks(image, layout):
 def extract_key_values(blocks):
     data = {}
     for block in blocks:
-        text = block["text"]
-        if "invoice" in text.lower():
-            data["invoice_number"] = text
-        elif "total" in text.lower():
-            data["total_amount"] = text
-        elif "customer" in text.lower():
-            data["customer_name"] = text
+        text = block["text"].lower()
+        if "invoice" in text:
+            data["Invoice Number"] = block["text"]
+        elif "total" in text:
+            data["Total Amount"] = block["text"]
+        elif "customer" in text:
+            data["Customer Name"] = block["text"]
     return data

requirements.txt

@@ -1,10 +1,8 @@
-layoutparser[layoutmodels_detectron2] @ git+https://github.com/Layout-Parser/layout-parser.git
-git+https://github.com/facebookresearch/detectron2.git
 streamlit
+layoutparser
 pdf2image
 pytesseract
 transformers
 torch
 Pillow
 opencv-python
-