Upload 2 files

app.py

@@ -1,52 +1,44 @@
 import streamlit as st
+from transformers import AutoFeatureExtractor, AutoModelForImageClassification
 from PIL import Image
 import torch
-from transformers import AutoImageProcessor, ConvNextForImageClassification
-import pandas as pd
-from utils.matcher import fuzzy_match
+from utils.matcher import oversæt_fuzzy
 
 @st.cache_resource
 def load_model():
-    processor = AutoImageProcessor.from_pretrained("duongna/convnext-tiny-food101")
-    model = ConvNextForImageClassification.from_pretrained("duongna/convnext-tiny-food101")
-    return processor, model
+    extractor = AutoFeatureExtractor.from_pretrained("nateraw/food101")
+    model = AutoModelForImageClassification.from_pretrained("nateraw/food101")
+    return extractor, model
 
-processor, model = load_model()
-data = pd.read_csv("kaloriedata.csv")
-madliste = data["navn"].tolist()
+extractor, model = load_model()
 
-st.title("🍽️ WebKalorier – Madanalyse")
+st.title("Kalorieestimering fra billede")
 
-uploaded = st.file_uploader("Upload et billede", type=["jpg", "jpeg", "png"])
+uploaded_file = st.file_uploader("Upload billede", type=["jpg", "png", "jpeg"])
+if uploaded_file:
+    image = Image.open(uploaded_file).convert("RGB")
+    st.image(image, caption="Uploadet billede", use_column_width=True)
 
-if uploaded:
-    img = Image.open(uploaded).convert("RGB")
-    st.image(img, caption="Uploadet billede", use_container_width=True)
-
-    inputs = processor(images=img, return_tensors="pt")
+    inputs = extractor(images=image, return_tensors="pt")
     with torch.no_grad():
-        logits = model(**inputs).logits
-    probs = torch.nn.functional.softmax(logits, dim=-1)
-    score, class_id = torch.max(probs, dim=1)
-    label = model.config.id2label[class_id.item()]
-    confidence = score.item()
+        outputs = model(**inputs)
+        probs = torch.nn.functional.softmax(outputs.logits, dim=1)
+        confidence, predicted_class = torch.max(probs, dim=1)
 
-    st.markdown(f"🤖 Modelgæt: `{label}` med {confidence:.0%} sikkerhed")
+    label = model.config.id2label[predicted_class.item()]
+    confidence_pct = confidence.item() * 100
 
-    if confidence < 0.7:
-        valgt = st.selectbox("Vælg fødevare manuelt:", madliste)
+    if confidence_pct > 70:
+        st.subheader(f"Modelgæt: {label} ({confidence_pct:.1f}%)")
     else:
-        valgt = fuzzy_match(label, madliste)
-
-    gram = st.number_input("Estimeret mængde (g):", 1, 1000, 150)
-    række = data[data["navn"] == valgt]
-    if not række.empty:
-        kcal100 = række["kcal_pr_100g"].values[0]
-        samlet = round(kcal100 * gram / 100)
-        st.success(f"{gram} g {valgt} = {samlet} kcal")
-
-    feedback = st.text_input("Feedback / korrektion:")
-    if st.button("Send feedback"):
-        with open("feedback_log.txt", "a") as f:
-            f.write(f"{label},{valgt},{confidence:.2f},{feedback}\n")
-        st.info("Tak for din feedback! 🙏")
+        label = st.selectbox("Model usikker – vælg selv fødevare:", list(model.config.id2label.values()))
+        st.subheader(f"Valgt: {label}")
+
+    # Fuzzy matching
+    dansk_label = oversæt_fuzzy(label)
+    st.write(f"Fortolket som: {dansk_label}")
+
+    # Feedback
+    feedback = st.selectbox("Passede forslaget?", ["Ja", "Nej"])
+    st.write("Tak for feedback!")
+

utils/matcher.py

@@ -1,6 +1,15 @@
-from fuzzywuzzy import process
+from difflib import get_close_matches
 
-def fuzzy_match(label, liste):
-    label = label.replace("_", " ").lower()
-    match, score = process.extractOne(label, liste)
-    return match if score > 70 else label
+# Simple mapping from English label to Danish synonym
+mapping = {
+    "french_fries": "pommes frites",
+    "pizza": "pizza",
+    "ice_cream": "is",
+    "apple_pie": "æbletærte",
+    "grilled_salmon": "grillet laks"
+    # ... tilføj flere som ønsket
+}
+
+def oversæt_fuzzy(label):
+    match = get_close_matches(label, mapping.keys(), n=1, cutoff=0.5)
+    return mapping[match[0]] if match else label