streamlit_app.py

# ✅ Fix Hugging Face Space permission error
import os
os.environ["STREAMLIT_HOME"] = "/tmp"
os.environ["STREAMLIT_BROWSER_GATHER_USAGE_STATS"] = "false"

import streamlit as st
from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import numpy as np
from PIL import Image

# ✅ Load model from Hugging Face Model Hub
@st.cache_resource
def load_model_from_hf():
    model_path = hf_hub_download(
        repo_id="1Codephoenix/fish-freshness-model",
        filename="fish_freshness_model_retrained_final.keras"
    )
    return load_model(model_path)

model = load_model_from_hf()

# ✅ Class labels and messages
class_names = ['Fresh', 'Moderately Fresh', 'Spoiled']
custom_messages = {
    'Fresh': (
        "✅ **Fresh Fish Detected**\n"
        "- Age: Less than 1 day\n"
        "- Bright eyes, red gills, firm flesh\n"
        "- Safe to eat raw or cooked"
    ),
    'Moderately Fresh': (
        "⚠️ **Moderately Fresh**\n"
        "- Age: 2–3 days\n"
        "- Slight odor, softer texture\n"
        "- Should be cooked thoroughly"
    ),
    'Spoiled': (
        "🚫 **Spoiled or Unsafe Fish**\n"
        "- Age: 4+ days or preserved\n"
        "- Strong odor, dull eyes, mushy flesh\n"
        "- Unsafe to consume"
    )
}

# ✅ Streamlit UI
st.set_page_config(page_title="Fish Freshness Classifier", page_icon="🐟")
st.title("🐟 AI-Powered Fish Freshness Classifier")
st.subheader("Upload a fish image to predict its freshness level")

uploaded_file = st.file_uploader("Choose a fish image", type=["jpg", "jpeg", "png"])

if uploaded_file:
    try:
        # Preprocess image
        img = Image.open(uploaded_file).convert("RGB")
        img_resized = img.resize((224, 224))
        img_array = image.img_to_array(img_resized)
        img_array = np.expand_dims(img_array / 255.0, axis=0)

        # Predict
        prediction = model.predict(img_array)
        predicted_index = np.argmax(prediction)
        predicted_class = class_names[predicted_index]
        confidence = prediction[0][predicted_index]

        # Display
        st.image(img, caption="Uploaded Fish Image", use_column_width=True)
        st.markdown(f"### 🎯 Prediction: **{predicted_class}** ({confidence * 100:.2f}%)")
        st.markdown(custom_messages[predicted_class])

        # Probabilities
        st.subheader("📊 Prediction Confidence:")
        for i, label in enumerate(class_names):
            st.write(f"- {label}: {prediction[0][i] * 100:.2f}%")

    except Exception as e:
        st.error(f"❌ Error processing image: {e}")