feat: added more result and modified layout

README.md

@@ -1,5 +1,5 @@
 ---
-title: Layoutlm Invoice
+title: LayoutLM Invoice
 emoji: 📚
 colorFrom: pink
 colorTo: blue
@@ -9,4 +9,16 @@ app_file: app.py
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+## LayoutLM for Invoice
+
+### Pre-requirements
+
+Tesseract requires you to install `tesseract-ocr` via `apt` command, just adding it in `packages.txt`.
+
+```text
+tesseract-ocr
+```
+
+### Secure
+
+This service is private application, requires user name and password (ref. secret variable `USERNAME` & `PASSWORD`).

apis/layoutlm.py

@@ -41,21 +41,25 @@ class LayoutLM:
         return answer
 
     def inference(self, img, is_debug=False):
-        merchant_id = self.answer_the_question(img, 'What is merchant ID?')
-        merchant_name = self.answer_the_question(img, 'What is merchant name?')
-        merchant_address = self.answer_the_question(img, 'What is merchant address?')
-        invoice_no = self.answer_the_question(img, 'What is tax invoice number?')
-        products = self.answer_the_question(img, 'What are buy products?')
-        pos_no = self.answer_the_question(img, 'What is POS number?')
-        net_price = self.answer_the_question(img, 'What is the net-price?')
-        date_time = self.answer_the_question(img, 'What date, year and time of the invoice?')
+        merchant_id = self.answer_the_question(img, 'What is merchant ID?', is_debug=is_debug)
+        merchant_name = self.answer_the_question(img, 'What is merchant name?', is_debug=is_debug)
+        merchant_address = self.answer_the_question(img, 'What is merchant address?', is_debug=is_debug)
+        merchant_branch = self.answer_the_question(img, 'What is branch of merchant?', is_debug=is_debug)
+        invoice_no = self.answer_the_question(img, 'What is invoice number or INV?', is_debug=is_debug)
+        products = self.answer_the_question(img, 'What are buy products?', is_debug=is_debug)
+        product_codes = self.answer_the_question(img, 'What are code of buy products?', is_debug=is_debug)
+        pos_no = self.answer_the_question(img, 'What is POS number?', is_debug=is_debug)
+        net_price = self.answer_the_question(img, 'What is the net-price?', is_debug=is_debug)
+        date_time = self.answer_the_question(img, 'What date, year and time of the invoice?', is_debug=is_debug)
 
         if is_debug:
             print(f'Merchant ID: {merchant_id}')
             print(f'Merchant name: {merchant_name}')
             print(f'Merchant address: {merchant_address}')
+            print(f'Merchant branch: {merchant_branch}')
             print(f'Invoice no.: {invoice_no}')
             print(f'Products: {products}')
+            print(f'Product codes: {product_codes}')
             print(f'POS no.: {pos_no}')
             print(f'Net price: {net_price}')
             print(f'Date/Time: {date_time}')
@@ -65,8 +69,10 @@ class LayoutLM:
                 'Merchant ID', 
                 'Merchant name', 
                 'Merchant address', 
+                'Merchant branch', 
                 'Invoice no.',
                 'Products',
+                'Product codes',
                 'POS no.',
                 'Net price',
                 'Date/Time'
@@ -75,8 +81,10 @@ class LayoutLM:
                 str(merchant_id),
                 str(merchant_name),
                 str(merchant_address),
+                str(merchant_branch),
                 str(invoice_no),
                 str(products),
+                str(product_codes),
                 str(pos_no),
                 str(net_price),
                 str(date_time)

app.py

@@ -13,25 +13,36 @@ def auth(username, password):
 def inference(img) -> pd.DataFrame:
     return layoutlm.inference(img)
 
-if __name__ == "__main__":
+def ask(img, question) -> str:
+    return layoutlm.answer_the_question(img, question, is_debug=True)
 
-    try:
-        layoutlm = LayoutLM()
-        layoutlm.set_model(layoutlm.default_model)
+with gr.Blocks() as demo:
 
-        demo = gr.Interface(
-            inference,
-            gr.Image(type="pil"),
-            gr.Dataframe(
+    with gr.Column():
+        inp_img = gr.Image(type='pil')
+    
+        with gr.Row():
+            inp_img = gr.Image(type='pil')
+            out = gr.Dataframe(
                 headers=['Data', 'Value'],
                 datatype=['str', 'str'],
                 row_count=8,
                 col_count=(2, 'fixed'), 
                 interactive=False
             )
-        )
 
-        demo.launch(auth=auth)
+            txt_custom_question = gr.Textbox(label='Your question')
+            btn_ask = gr.Button('Ask me')
+            txt_out_answer = gr.Textbox(label='Answer', interactive=False)
+
+            # event
+            inp_img.change(inference, inp_img, out)
+            btn_ask.click(ask, [inp_img, txt_custom_question], txt_out_answer)
 
+if __name__ == '__main__':
+    try:
+        layoutlm = LayoutLM()
+        layoutlm.set_model(layoutlm.default_model)
+        demo.launch(auth=auth)
     except Exception as e:
         print(str(e))