import streamlit as st
import os
from PIL import Image
import torch
from torchvision import transforms
from models.cnn import CNNModel
from utils.transforms import get_transforms

os.environ["STREAMLIT_ROOT"] = "/tmp/.streamlit"

@st.cache_resource
def load_model(model_path='saved_models/cnn_model.pth'):
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    checkpoint = torch.load(model_path, map_location=device)
    class_names = checkpoint['class_names']
    model = CNNModel(num_classes=len(class_names))
    model.load_state_dict(checkpoint['model_state_dict'])
    model.to(device)
    model.eval()
    return model, class_names, device

st.title("📸 Intel Image Classification")
st.write("Upload an image to classify it into one of the image categories: buildings, forest, glacier, mountain, sea, or street.")

model, class_names, device = load_model()

uploaded_file = st.file_uploader("Upload an image", type=["jpg", "jpeg", "png"])

if uploaded_file:
    image = Image.open(uploaded_file).convert("RGB")
    st.image(image, caption="Uploaded Image", use_container_width=True)

    transform = get_transforms(train=False)
    image_tensor = transform(image).unsqueeze(0).to(device)

    with torch.no_grad():
        output = model(image_tensor)
        predicted_idx = torch.argmax(output, 1).item()
        predicted_class = class_names[predicted_idx]

    st.success(f"Predicted class: {predicted_class}")