improve demo

app.py

@@ -1,56 +1,7 @@
-import re
-
-from annotated_text import annotated_text
 import requests
 import streamlit as st
-
-
-BARCODE_PATH_REGEX = re.compile(r"^(...)(...)(...)(.*)$")
-
-
-def split_barcode(barcode: str) -> list[str]:
-    """Split barcode in the same way as done by Product Opener to generate a
-    product image folder.
-
-    :param barcode: The barcode of the product. For the pro platform only,
-        it must be prefixed with the org ID using the format
-        `{ORG_ID}/{BARCODE}`
-    :raises ValueError: raise a ValueError if `barcode` is invalid
-    :return: a list containing the splitted barcode
-    """
-    org_id = None
-    if "/" in barcode:
-        # For the pro platform, `barcode` is expected to be in the format
-        # `{ORG_ID}/{BARCODE}` (ex: `org-lea-nature/3307130803004`)
-        org_id, barcode = barcode.split("/", maxsplit=1)
-
-    if not barcode.isdigit():
-        raise ValueError(f"unknown barcode format: {barcode}")
-
-    match = BARCODE_PATH_REGEX.fullmatch(barcode)
-
-    splits = [x for x in match.groups() if x] if match else [barcode]
-
-    if org_id is not None:
-        # For the pro platform only, images and OCRs belonging to an org
-        # are stored in a folder named after the org for all its products, ex:
-        # https://images.pro.openfoodfacts.org/images/products/org-lea-nature/330/713/080/3004/1.jpg
-        splits.append(org_id)
-
-    return splits
-
-
-def _generate_file_path(barcode: str, image_id: str, suffix: str):
-    splitted_barcode = split_barcode(barcode)
-    return f"/{'/'.join(splitted_barcode)}/{image_id}{suffix}"
-
-
-def generate_ocr_path(barcode: str, image_id: str) -> str:
-    return _generate_file_path(barcode, image_id, ".json")
-
-
-def generate_image_path(barcode: str, image_id: str) -> str:
-    return _generate_file_path(barcode, image_id, ".400.jpg")
+from annotated_text import annotated_text
+from openfoodfacts.images import generate_image_url, generate_json_ocr_url
 
 
 @st.cache_data
@@ -85,7 +36,11 @@ def display_ner_tags(text: str, entities: list[dict]):
     annotated_text(spans)
 
 
-def run(barcode: str, model_version: str, min_threshold: float = 0.5,):
+def run(
+    barcode: str,
+    model_version: str,
+    min_threshold: float = 0.5,
+):
     product = get_product(barcode)
     st.markdown(f"[Product page](https://world.openfoodfacts.org/product/{barcode})")
 
@@ -93,18 +48,21 @@ def run(barcode: str, model_version: str, min_threshold: float = 0.5,):
         st.error(f"Product {barcode} not found")
         return
 
-    images = product["images"]
+    images = product.get("images", [])
+
+    if not images:
+        st.error(f"No images found for product {barcode}")
+        return
+
     for image_id, _ in images.items():
         if not image_id.isdigit():
             continue
 
-        ocr_path = generate_ocr_path(barcode, image_id)
-        ocr_url = f"https://static.openfoodfacts.org/images/products{ocr_path}"
+        ocr_url = generate_json_ocr_url(barcode, image_id)
         prediction = send_prediction_request(ocr_url, model_version)
 
         st.divider()
-        image_path = generate_image_path(barcode, image_id)
-        image_url = f"https://static.openfoodfacts.org/images/products{image_path}"
+        image_url = generate_image_url(barcode, image_id)
         st.markdown(f"[Image {image_id}]({image_url}), [OCR]({ocr_url})")
         st.image(image_url)
 
@@ -126,7 +84,9 @@ st.markdown(
     "This demo leverages the ingredient entity detection model, "
     "that takes the OCR text as input and predict ingredient lists."
 )
-barcode = st.text_input("barcode", help="Barcode of the product", value=default_barcode).strip()
+barcode = st.text_input(
+    "barcode", help="Barcode of the product", value=default_barcode
+).strip()
 model_version = "1"
 st.experimental_set_query_params(barcode=barcode)
 

requirements.txt

@@ -1,3 +1,4 @@
 requests==2.28.1
-streamlit==1.15.1
-st-annotated-text==4.0.0
+streamlit==1.28.1
+st-annotated-text==4.0.0
+openfoodfacts==0.1.11