add app

app.py

@@ -0,0 +1,133 @@
+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")
+
+
+@st.cache_data
+def send_prediction_request(ocr_url: str):
+    return requests.get(
+        "https://robotoff.openfoodfacts.net/api/v1/predict/ingredient_list",
+        params={"ocr_url": ocr_url},
+    ).json()
+
+
+def get_product(barcode: str):
+    r = requests.get(f"https://world.openfoodfacts.org/api/v2/product/{barcode}")
+
+    if r.status_code == 404:
+        return None
+
+    return r.json()["product"]
+
+
+def display_ner_tags(text: str, entities: list[dict]):
+    spans = []
+    previous_idx = 0
+    for entity in entities:
+        score = entity["score"]
+        lang = entity["lang"]["lang"]
+        start_idx = entity["start"]
+        end_idx = entity["end"]
+        spans.append(text[previous_idx:start_idx])
+        spans.append((text[start_idx:end_idx], f"score={score:.3f} | lang={lang}"))
+        previous_idx = end_idx
+    spans.append(text[previous_idx:])
+    annotated_text(spans)
+
+
+def run(barcode: str, min_threshold: float = 0.5):
+    product = get_product(barcode)
+
+    if not product:
+        st.error(f"Product {barcode} not found")
+        return
+
+    images = product["images"]
+    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}"
+        prediction = send_prediction_request(ocr_url)
+
+        entities = prediction["entities"]
+        text = prediction["text"]
+        filtered_entities = [e for e in entities if e["score"] >= min_threshold]
+
+        if filtered_entities:
+            st.divider()
+            image_path = generate_image_path(barcode, image_id)
+            image_url = f"https://static.openfoodfacts.org/images/products{image_path}"
+            st.image(image_url)
+            display_ner_tags(text, filtered_entities)
+
+
+query_params = st.experimental_get_query_params()
+default_barcode = query_params["barcode"][0] if "barcode" in query_params else ""
+
+st.title("Ingredient extraction demo")
+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)
+threshold = st.number_input(
+    "threshold",
+    help="Minimum threshold for entity predictions",
+    min_value=0.0,
+    max_value=1.0,
+    value=0.98,
+)
+
+if barcode:
+    run(barcode, threshold)

requirements.txt

@@ -0,0 +1,3 @@
+requests==2.28.1
+streamlit==1.15.1
+st-annotated-text==4.0.0