How to detect objects in real time with a cv2 image?

[shershunov]

I want to detect objects in a cv2 image in real time. I have a code that captures the screen using the mfc library. How do I send an image to yolov5 at maximum speed? If you can help me, I will be very grateful!

from cv2 import imshow, waitKey, destroyAllWindows
from mss import mss
from numpy import array

with mss() as sct:
    monitor = {"top": 0, "left": 0, "width": 600, "height": 600}
    while True:
        img = array(sct.grab(monitor))
        imshow("Detect", img)
        if waitKey(1) & 0xFF == ord("q"):
            destroyAllWindows()
            break

[glenn-jocher]

@HamkiWhite you can use YOLOv5 PyTorch Hub models for inference in your loop.

PyTorch Hub

Inference with YOLOv5 and PyTorch Hub:

import torch

# Model
model = torch.hub.load('ultralytics/yolov5', 'yolov5s')

# Image
img = 'https://github.com/ultralytics/yolov5/raw/master/data/images/zidane.jpg'

# Inference
results = model(img)

[shershunov]

Thank)

[9christian9]

if you have success, share the code that always is good

[shershunov]

Ok)

[shershunov]

img = array(sct.grab(dimensions))
        result = model(img, size=640)
        repos = result.xyxy[0]
        if len(repos) > 0:
            a = 0
            for i in range(len(repos)):
                pos = torch.tensor(repos)[a]
                rectangle(img, (int(pos[0]), int(pos[1])), (int(pos[2]), int(pos[3])), (255, 255, 255), 2)
                a += 1
        imshow('Detect', img)

[9christian9]

@EkelviNistars
I've seen that .detach().clone() is the best practice for clone a tensor
I've removed the variable a to control a cycle... try to use i inside the cycle
I suggest changing your code in:

    result = model(img, size=640)
    repos = result.xyxy[0]
    if len(repos) > 0:
        for i in range(len(repos)):
            pos = repos[i].detach().clone()
            rectangle(img, (int(pos[0]), int(pos[1])), (int(pos[2]), int(pos[3])), (255, 255, 255), 2)
    imshow('Detect', img)