0.20.0/how_to/detect_and_annotate/

[giscus[bot]]

0.20.0/how_to/detect_and_annotate/

A set of easy-to-use utilities that will come in handy in any computer vision project.

https://supervision.roboflow.com/0.20.0/how_to/detect_and_annotate/

[s-madhavan001]

supervision can detect the objects in realtime using webcam ?
like "cv2.VideoCapture(0)" or ultralytics "source = 0"

its can using web cam to detect the object?

[SkalskiP]

Sure. You just need to read the frame from the webcam.

import cv2

camera = cv2.VideoCapture(0)
while True:
    _, image = camera.read()

Than pass that frame to model. For inference package the code snippet would look like this.

import cv2
import supervision as sv
from inference import get_model

model = get_model(model_id="yolov8n-640")

camera = cv2.VideoCapture(0)
while True:
    _, image = camera.read()
    results = model.infer(image)[0]
    detections = sv.Detections.from_inference(results)

[s-madhavan001]

its work but i cant show the detection

help me in show this detection result visualization

[SkalskiP]

You need to use cv2.imshow to display images/videos on your screen.

import cv2
import supervision as sv
from inference import get_model

model = get_model(model_id="yolov8n-640")
bounding_box_annotator = sv.BoundingBoxAnnotator()
label_annotator = sv.LabelAnnotator()

camera = cv2.VideoCapture(0)
while True:
    _, image = camera.read()
    results = model.infer(image)[0]
    detections = sv.Detections.from_inference(results)

    annotated_image = bounding_box_annotator.annotate(
        scene=image, detections=detections)
    annotated_image = label_annotator.annotate(
        scene=annotated_image, detections=detections)

    cv2.imshow("Processed Video", annotated_frame)
    if cv2.waitKey(1) & 0xFF == ord("q"):
        break
cv2.destroyAllWindows()

[devyoudini]

How do i integrate this into tkinter? anyone can help?

[LinasKo]

Hi @devyoudini 👋

I don't think we have instructions for that. However, with supervision you can obtain an annotated image which you should be able to display in tkinter exactly like an image from any other source.

[SkalskiP]

Hi @devyoudini 👋🏻 I made this short demo script.

import tkinter as tk
import supervision as sv
import numpy as np
from PIL import Image, ImageTk

IMAGE_PATH = "..."
BOUNDING_BOX_ANNOTATOR = sv.BoundingBoxAnnotator()
LABEL_ANNOTATOR = sv.LabelAnnotator()


def run():
    root = tk.Tk()
    root.title("app")

    # detections should come from AI model output
    detections = sv.Detections(
        xyxy=np.array([
            [0, 0, 100, 100]
        ]),
        class_id=np.array([0]),
    )
    
    image = Image.open(IMAGE_PATH)
    annotated_image = BOUNDING_BOX_ANNOTATOR.annotate(
        scene=image, detections=detections)
    annotated_image = LABEL_ANNOTATOR.annotate(
        scene=annotated_image, detections=detections)

    tk_image = ImageTk.PhotoImage(annotated_image)

    label = tk.Label(root, image=tk_image)
    label.pack()

    root.mainloop()


if __name__ == "__main__":
    run()

[vck]

can I use this library on SBCs that don't have access to CUDA/Nvidia GPUs?

[vck]

#699
Just found a discussion on supervision repo, it appear to be working WELL on raspberry pi.

[SkalskiP]

Yup. It should run just fine.