app.py

from typing import Tuple

import gradio as gr
import numpy as np
import supervision as sv
from inference import get_model

MARKDOWN = """
<h1 style='text-align: center'>YOLO-ARENA 🏟️</h1>

Welcome to YOLO-Arena! This demo showcases the performance of top object detection models pre-trained on the COCO dataset.

Powered by [inference](https://github.com/roboflow/inference) and [supervision](https://github.com/roboflow/supervision). 💜
"""

IMAGE_EXAMPLES = [
    ['https://media.roboflow.com/supervision/image-examples/people-walking.png', 0.3, 0.3, 0.1, 0.3, 0.3, 0.5],
    ['https://media.roboflow.com/supervision/image-examples/vehicles.png', 0.3, 0.3, 0.1, 0.3, 0.4, 0.5],
    ['https://media.roboflow.com/supervision/image-examples/basketball-1.png', 0.3, 0.3, 0.1, 0.3, 0.4, 0.5],
]

YOLO_V8_MODEL = get_model(model_id="coco/8")
YOLO_V9_MODEL = get_model(model_id="coco/17")
YOLO_V10_MODEL = get_model(model_id="coco/22")
YOLO11_MODEL = get_model(model_id="coco/27")
RF_DETR_MODEL = get_model(model_id="coco/36")

LABEL_ANNOTATORS = sv.LabelAnnotator(text_color=sv.Color.BLACK)
BOUNDING_BOX_ANNOTATORS = sv.BoxAnnotator()


def detect_and_annotate(
    model,
    input_image: np.ndarray,
    confidence_threshold: float,
    iou_threshold: float,
    class_id_mapping: dict = None
) -> np.ndarray:
    result = model.infer(
        input_image,
        confidence=confidence_threshold,
        iou_threshold=iou_threshold
    )[0]
    detections = sv.Detections.from_inference(result)

    if class_id_mapping:
        detections.class_id = np.array([
            class_id_mapping[class_id]
            for class_id
            in detections.class_id
        ])

    labels = [
        f"{class_name} ({confidence:.2f})"
        for class_name, confidence
        in zip(detections['class_name'], detections.confidence)
    ]

    annotated_image = input_image.copy()
    annotated_image = BOUNDING_BOX_ANNOTATORS.annotate(
        scene=annotated_image, detections=detections)
    annotated_image = LABEL_ANNOTATORS.annotate(
        scene=annotated_image, detections=detections, labels=labels)
    return annotated_image


def process_image(
    input_image: np.ndarray,
    yolo_v8_confidence_threshold: float,
    yolo_v9_confidence_threshold: float,
    yolo_v10_confidence_threshold: float,
    yolo11_confidence_threshold: float,
    rf_detr_confidence_threshold: float,
    iou_threshold: float
) -> Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
    yolo_v8_annotated_image = detect_and_annotate(
        YOLO_V8_MODEL, input_image, yolo_v8_confidence_threshold, iou_threshold)
    yolo_v9_annotated_image = detect_and_annotate(
        YOLO_V9_MODEL, input_image, yolo_v9_confidence_threshold, iou_threshold)
    yolo_10_annotated_image = detect_and_annotate(
        YOLO_V10_MODEL, input_image, yolo_v10_confidence_threshold, iou_threshold)
    yolo11_annotated_image = detect_and_annotate(
        YOLO11_MODEL, input_image, yolo11_confidence_threshold, iou_threshold)
    rf_detr_annotated_image = detect_and_annotate(
        RF_DETR_MODEL, input_image, rf_detr_confidence_threshold, iou_threshold)

    return (
        yolo_v8_annotated_image,
        yolo_v9_annotated_image,
        yolo_10_annotated_image,
        yolo11_annotated_image,
        rf_detr_annotated_image
    )


yolo_v8_confidence_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.3,
    step=0.01,
    label="YOLOv8 Confidence Threshold",
)

yolo_v9_confidence_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.3,
    step=0.01,
    label="YOLOv9 Confidence Threshold",
)

yolo_v10_confidence_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.3,
    step=0.01,
    label="YOLOv10 Confidence Threshold",
)

yolo11_confidence_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.3,
    step=0.01,
    label="YOLO11 Confidence Threshold",
)
rf_detr_confidence_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.4,
    step=0.01,
    label="RF-DETR Confidence Threshold",
)

iou_threshold_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.5,
    step=0.01,
    label="IoU Threshold",
    info=(
        "The Intersection over Union (IoU) threshold for non-maximum suppression. "
        "Decrease the value to lessen the occurrence of overlapping bounding boxes, "
        "making the detection process stricter. On the other hand, increase the value "
        "to allow more overlapping bounding boxes, accommodating a broader range of "
        "detections."
    ))


with gr.Blocks() as demo:
    gr.Markdown(MARKDOWN)
    with gr.Accordion("Configuration", open=False):
        with gr.Row():
            iou_threshold_component.render()
            yolo_v8_confidence_threshold_component.render()
            yolo_v9_confidence_threshold_component.render()
        with gr.Row():
            yolo_v10_confidence_threshold_component.render()
            yolo11_confidence_threshold_component.render()
            rf_detr_confidence_threshold_component.render()
    with gr.Row():
        input_image_component = gr.Image(
            type='pil',
            label='Input'
        )
        yolo_v8_output_image_component = gr.Image(
            type='pil',
            label='YOLOv8'
        )
        yolo_v9_output_image_component = gr.Image(
            type='pil',
            label='YOLOv9'
        )
    with gr.Row():
        yolo_v10_output_image_component = gr.Image(
            type='pil',
            label='YOLOv10'
        )
        yolo11_output_image_component = gr.Image(
            type='pil',
            label='YOLO11'
        )
        rf_detr_output_image_component = gr.Image(
            type='pil',
            label='RF-DETR'
        )
    submit_button_component = gr.Button(
        value='Submit',
        scale=1,
        variant='primary'
    )
    gr.Examples(
        fn=process_image,
        examples=IMAGE_EXAMPLES,
        inputs=[
            input_image_component,
            yolo_v8_confidence_threshold_component,
            yolo_v9_confidence_threshold_component,
            yolo_v10_confidence_threshold_component,
            yolo11_confidence_threshold_component,
            rf_detr_confidence_threshold_component,
            iou_threshold_component
        ],
        outputs=[
            yolo_v8_output_image_component,
            yolo_v9_output_image_component,
            yolo_v10_output_image_component,
            yolo11_output_image_component,
            rf_detr_output_image_component
        ],
        cache_examples=True,
        run_on_click=True
    )

    submit_button_component.click(
        fn=process_image,
        inputs=[
            input_image_component,
            yolo_v8_confidence_threshold_component,
            yolo_v9_confidence_threshold_component,
            yolo_v10_confidence_threshold_component,
            yolo11_confidence_threshold_component,
            rf_detr_confidence_threshold_component,
            iou_threshold_component
        ],
        outputs=[
            yolo_v8_output_image_component,
            yolo_v9_output_image_component,
            yolo_v10_output_image_component,
            yolo11_output_image_component,
            rf_detr_output_image_component
        ]
    )

demo.launch(debug=False, show_error=True, max_threads=1)