src/streamlit_app.py

import streamlit as st
from ultralytics import YOLO
from PIL import Image
import io

# --- Page Configuration ---
st.set_page_config(
    page_title="YOLO Object Detection For Satellite Image",
    page_icon="🤖",
    layout="wide",
    initial_sidebar_state="expanded",
)

# --- Theme-Aware Custom CSS ---
st.markdown("""
<style>
    /* --- Base Styles (Shared) --- */
    .st-expander {
        border-radius: 10px;
    }
    /* --- LIGHT THEME --- */
    [data-theme="light"] .stApp {
        background-color: #f0f2f6; /* Light gray background */
    }
    [data-theme="light"] h1 {
        color: #1E3A8A; /* Deep blue */
    }
    [data-theme="light"] h2, [data-theme="light"] h3 {
        color: #3B82F6; /* Lighter blue */
    }
    [data-theme="light"] .st-expander {
        border: 1px solid #ddd;
        background-color: #ffffff;
    }
    /* --- DARK THEME --- */
    [data-theme="dark"] .stApp {
        background-color: #0E1117; /* Default Streamlit dark background */
    }
    [data-theme="dark"] h1 {
        color: #60A5FA; /* A nice light blue for titles */
    }
    [data-theme="dark"] h2, [data-theme="dark"] h3 {
        color: #93C5FD; /* A slightly lighter blue */
    }
    [data-theme="dark"] .st-expander {
        border: 1px solid #444;
        background-color: #1a1c24; /* A slightly lighter dark shade */
    }
    /* Ensure default text is visible in dark mode */
    [data-theme="dark"] body {
        color: #FAFAFA;
    }
</style>
""", unsafe_allow_html=True)

# --- Model Loading ---
@st.cache_resource
def load_model(model_path):
    """
    Loads the YOLO model from the specified path.
    Caches the model to avoid reloading on every interaction.
    """
    try:
        model = YOLO(model_path)
        return model
    except Exception as e:
        st.error(f"Error loading model: {e}")
        return None

# Path to the model file inside the 'src' directory
MODEL_PATH = './src/rssi_last.pt'
model = load_model(MODEL_PATH)

# --- Sidebar ---
st.sidebar.header("Configuration")
confidence_threshold = st.sidebar.slider(
    "Confidence Threshold", 0.0, 1.0, 0.4, 0.05
)
st.sidebar.markdown("---")
uploaded_file = st.sidebar.file_uploader(
    "Upload an image...", type=["jpg", "jpeg", "png"]
)
st.sidebar.markdown("---")
st.sidebar.markdown(
    "**About this App**\n\n"
    "This application uses a custom-trained YOLO model to detect objects in images. "
    "Upload an image and see the magic!"
)

# --- Main Page ---
st.title("🖼️ Custom Object Detection with YOLO for Satellite image")

if uploaded_file is not None:
    # Read the uploaded image file
    image_data = uploaded_file.getvalue()
    original_image = Image.open(io.BytesIO(image_data))
    
    # Create two columns for side-by-side display
    col1, col2 = st.columns(2)
    
    with col1:
        st.subheader("Original Image")
        # CHANGED: use_container_width is the new, recommended parameter
        st.image(original_image, caption="Your uploaded image.", use_container_width=True)

    if model:
        # Perform inference
        with st.spinner("Running detection..."):
            results = model(original_image, conf=confidence_threshold)
        
        # The result object contains the annotated image and detection data
        result = results[0]
        
        # Use the plot() method to get an annotated image (in BGR format)
        annotated_image_bgr = result.plot()
        # Convert BGR to RGB for display in Streamlit
        annotated_image_rgb = annotated_image_bgr[..., ::-1]

        with col2:
            st.subheader("Detected Objects")
            # CHANGED: use_container_width is the new, recommended parameter
            st.image(annotated_image_rgb, caption="Image with detected objects.", use_container_width=True)

        # Display detection details
        st.subheader("Detection Details")
        if len(result.boxes) > 0:
            with st.expander("Click to see detailed results", expanded=True):
                # Extract details for each detected box
                for i, box in enumerate(result.boxes):
                    label = result.names[box.cls[0].item()]
                    conf = box.conf[0].item()
                    xywhn = box.xywhn[0].tolist() # Normalized xywh
                    
                    st.markdown(f"**Object {i+1}: `{label}`**")
                    st.write(f"- Confidence: **{conf:.2f}**")
                    st.write(f"- Bounding Box (Normalized xywh):")
                    st.code(f"  x: {xywhn[0]:.4f}, y: {xywhn[1]:.4f}, width: {xywhn[2]:.4f}, height: {xywhn[3]:.4f}")
        else:
            st.info("No objects were detected with the current confidence threshold.")
else:
    st.info("Please upload an image using the sidebar to begin.")