app.py

import streamlit as st
import cv2
import numpy as np
from PIL import Image
from ultralytics import YOLO
import os
import time
import matplotlib.pyplot as plt
from streamlit_webrtc import webrtc_streamer, WebRtcMode
import functools
import threading

# --- Page Configuration ---  MUST BE THE VERY FIRST STREAMLIT CALL
st.set_page_config(page_title="YOLOv11 Object Detection App", page_icon="🧊", layout="wide")

# --- Load YOLO Model ---
@st.cache_resource
def load_model():
    def load_in_thread():
        try:
            model = YOLO("yolov8n.pt")  # Or your custom model path
            return model
        except Exception as e:
            st.error(f"Error loading YOLO model: {e}")
            return None

    model_thread = threading.Thread(target=load_in_thread)
    model_thread.start()
    model_thread.join()
    return load_in_thread()

model = load_model()

# --- Functions ---
def detect_objects(image, confidence_threshold=0.5, iou_threshold=0.45):
    if model is None:
        st.warning("YOLOv11 model not loaded.")
        return []

    results = model(image, conf=confidence_threshold, iou=iou_threshold)
    detections = []
    for result in results:
        if result.boxes is not None and len(result.boxes) > 0:
            for box in result.boxes:
                confidence = float(box.conf[0])
                class_id = int(box.cls[0])
                class_name = model.names[class_id]
                xyxy = box.xyxy[0].tolist()
                x1, y1, x2, y2 = map(int, xyxy)
                detections.append(((x1, y1, x2, y2), class_name, confidence))
    return detections

def draw_boxes(image, detections):
    for (x1, y1, x2, y2), class_name, confidence in detections:
        label = f"{class_name}: {confidence:.2f}"
        cv2.rectangle(image, (x1, y1), (x2, y2), (0, 255, 0), 2)
        cv2.putText(image, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
    return image

def process_video(video_file, confidence_threshold, iou_threshold):
    try:
        video_bytes = video_file.read()
        video_path = "temp_video.mp4"
        with open(video_path, "wb") as temp_file:
            temp_file.write(video_bytes)

        cap = cv2.VideoCapture(video_path)
        frame_placeholder = st.empty()

        while cap.isOpened():
            ret, frame = cap.read()
            if not ret:
                break

            detections = detect_objects(frame, confidence_threshold, iou_threshold)
            frame_with_boxes = draw_boxes(frame.copy(), detections)

            frame_with_boxes_rgb = cv2.cvtColor(frame_with_boxes, cv2.COLOR_BGR2RGB)
            img = Image.fromarray(frame_with_boxes_rgb)

            frame_placeholder.image(img, caption="Processed Video", use_column_width=True)

        cap.release()
        os.remove(video_path)
    except Exception as e:
        st.error(f"Error processing video: {e}")

def process_frame(frame, confidence_threshold, iou_threshold):
    img = frame.to_ndarray(format="bgr24")
    detections = detect_objects(img, confidence_threshold, iou_threshold)
    img_with_boxes = draw_boxes(img.copy(), detections)
    return img_with_boxes

def simulate_fine_tuning(epochs=10):
    st.write("Simulating Fine-tuning...")  # Placeholder message
    time.sleep(2)  # Simulate some processing time
    st.success(f"Fine-tuning completed (simulated for {epochs} epochs).")


# --- Main App ---
def main():
    st.title("YOLOv11 Object Detection App")

    with st.sidebar:
        st.header("Settings")
        app_mode = st.selectbox("Choose the App mode", ["About", "Run on Image", "Run on Video", "Live Camera Feed", "Fine-Tune (Simulated)"])

        confidence_threshold = st.slider("Confidence Threshold", 0.0, 1.0, 0.5, 0.01, help="Min confidence for detected objects.")
        iou_threshold = st.slider("IoU Threshold", 0.0, 1.0, 0.45, 0.01, help="IoU threshold for NMS.")

        if app_mode == "Fine-Tune (Simulated)":
            num_epochs = st.slider("Number of Epochs", 1, 20, 5, 1)

    if app_mode == "About":
        st.subheader("About the App")
        st.write("This app uses YOLOv8n for object detection.")  # Updated model name
        st.write("Upload an image or video, or use your camera.")

    elif app_mode == "Run on Image":
        st.subheader("Object Detection on Image")
        uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

        if uploaded_file is not None:
            image = Image.open(uploaded_file)
            st.image(image, caption="Uploaded Image.", use_column_width=True)

            image_cv = np.array(image)
            image_cv = cv2.cvtColor(image_cv, cv2.COLOR_RGB2BGR)

            if st.button("Detect Objects"):
                detections = detect_objects(image_cv, confidence_threshold, iou_threshold)
                image_with_boxes = draw_boxes(image_cv.copy(), detections)
                image_with_boxes_rgb = cv2.cvtColor(image_with_boxes, cv2.COLOR_BGR2RGB)
                st.image(image_with_boxes_rgb, caption="Detected Objects", use_column_width=True)

    elif app_mode == "Run on Video":
        st.subheader("Object Detection on Video")
        video_file = st.file_uploader("Choose a video...", type=["mp4", "avi"])

        if video_file is not None:
            process_video(video_file, confidence_threshold, iou_threshold)

    elif app_mode == "Live Camera Feed":
        st.subheader("Live Camera Feed")
        custom_process_frame = functools.partial(process_frame, confidence_threshold=confidence_threshold, iou_threshold=iou_threshold)
        webrtc_streamer(key="live-feed", video_frame_callback=custom_process_frame, mode=WebRtcMode.SENDRECV, media_stream_constraints={"video": True, "audio": False})

    elif app_mode == "Fine-Tune (Simulated)":
        simulate_fine_tuning(epochs=num_epochs)

if __name__ == "__main__":
    main()