Reference for ultralytics/solutions/streamlit_inference.py

def inference(model=None):
    """Performs real-time object detection on video input using YOLO in a Streamlit web application."""
    check_requirements("streamlit>=1.29.0")  # scope imports for faster ultralytics package load speeds
    import streamlit as st

    from ultralytics import YOLO

    # Hide main menu style
    menu_style_cfg = """<style>MainMenu {visibility: hidden;}</style>"""

    # Main title of streamlit application
    main_title_cfg = """<div><h1 style="color:#FF64DA; text-align:center; font-size:40px; 
                             font-family: 'Archivo', sans-serif; margin-top:-50px;margin-bottom:20px;">
                    Ultralytics YOLO Streamlit Application
                    </h1></div>"""

    # Subtitle of streamlit application
    sub_title_cfg = """<div><h4 style="color:#042AFF; text-align:center; 
                    font-family: 'Archivo', sans-serif; margin-top:-15px; margin-bottom:50px;">
                    Experience real-time object detection on your webcam with the power of Ultralytics YOLO! 🚀</h4>
                    </div>"""

    # Set html page configuration
    st.set_page_config(page_title="Ultralytics Streamlit App", layout="wide", initial_sidebar_state="auto")

    # Append the custom HTML
    st.markdown(menu_style_cfg, unsafe_allow_html=True)
    st.markdown(main_title_cfg, unsafe_allow_html=True)
    st.markdown(sub_title_cfg, unsafe_allow_html=True)

    # Add ultralytics logo in sidebar
    with st.sidebar:
        logo = "https://raw.githubusercontent.com/ultralytics/assets/main/logo/Ultralytics_Logotype_Original.svg"
        st.image(logo, width=250)

    # Add elements to vertical setting menu
    st.sidebar.title("User Configuration")

    # Add video source selection dropdown
    source = st.sidebar.selectbox(
        "Video",
        ("webcam", "video"),
    )

    vid_file_name = ""
    if source == "video":
        vid_file = st.sidebar.file_uploader("Upload Video File", type=["mp4", "mov", "avi", "mkv"])
        if vid_file is not None:
            g = io.BytesIO(vid_file.read())  # BytesIO Object
            vid_location = "ultralytics.mp4"
            with open(vid_location, "wb") as out:  # Open temporary file as bytes
                out.write(g.read())  # Read bytes into file
            vid_file_name = "ultralytics.mp4"
    elif source == "webcam":
        vid_file_name = 0

    # Add dropdown menu for model selection
    available_models = [x.replace("yolo", "YOLO") for x in GITHUB_ASSETS_STEMS if x.startswith("yolo11")]
    if model:
        available_models.insert(0, model.split(".pt")[0])  # insert model without suffix as *.pt is added later

    selected_model = st.sidebar.selectbox("Model", available_models)
    with st.spinner("Model is downloading..."):
        model = YOLO(f"{selected_model.lower()}.pt")  # Load the YOLO model
        class_names = list(model.names.values())  # Convert dictionary to list of class names
    st.success("Model loaded successfully!")

    # Multiselect box with class names and get indices of selected classes
    selected_classes = st.sidebar.multiselect("Classes", class_names, default=class_names[:3])
    selected_ind = [class_names.index(option) for option in selected_classes]

    if not isinstance(selected_ind, list):  # Ensure selected_options is a list
        selected_ind = list(selected_ind)

    enable_trk = st.sidebar.radio("Enable Tracking", ("Yes", "No"))
    conf = float(st.sidebar.slider("Confidence Threshold", 0.0, 1.0, 0.25, 0.01))
    iou = float(st.sidebar.slider("IoU Threshold", 0.0, 1.0, 0.45, 0.01))

    col1, col2 = st.columns(2)
    org_frame = col1.empty()
    ann_frame = col2.empty()

    fps_display = st.sidebar.empty()  # Placeholder for FPS display

    if st.sidebar.button("Start"):
        videocapture = cv2.VideoCapture(vid_file_name)  # Capture the video

        if not videocapture.isOpened():
            st.error("Could not open webcam.")

        stop_button = st.button("Stop")  # Button to stop the inference

        while videocapture.isOpened():
            success, frame = videocapture.read()
            if not success:
                st.warning("Failed to read frame from webcam. Please make sure the webcam is connected properly.")
                break

            prev_time = time.time()  # Store initial time for FPS calculation

            # Store model predictions
            if enable_trk == "Yes":
                results = model.track(frame, conf=conf, iou=iou, classes=selected_ind, persist=True)
            else:
                results = model(frame, conf=conf, iou=iou, classes=selected_ind)
            annotated_frame = results[0].plot()  # Add annotations on frame

            # Calculate model FPS
            curr_time = time.time()
            fps = 1 / (curr_time - prev_time)

            # display frame
            org_frame.image(frame, channels="BGR")
            ann_frame.image(annotated_frame, channels="BGR")

            if stop_button:
                videocapture.release()  # Release the capture
                torch.cuda.empty_cache()  # Clear CUDA memory
                st.stop()  # Stop streamlit app

            # Display FPS in sidebar
            fps_display.metric("FPS", f"{fps:.2f}")

        # Release the capture
        videocapture.release()

    # Clear CUDA memory
    torch.cuda.empty_cache()

    # Destroy window
    cv2.destroyAllWindows()