Often when doing computer vision tasks, it's not the entire image or photograph you are interested in.
A photo of scene may be contain many different objects that need to be classified each. so how do you achieve that?
You can use region-of-interest. A region of the image that is important.
Lets say you want to recognize faces, given a photo a person, you don't need the hat or the clothes, you need a close up of the face.
In Python you can easily do that. If you load an image like this
#!/usr/bin/python3
import cv2
import numpy as np
img = cv2.imread("lena.png", cv2.IMREAD_UNCHANGED)
cv2.imshow("Demo", img)
Then you can create a region of interest like this
face = np.ones((200, 150, 3))
face = img[200:400, 200:350]
cv2.imshow("face", face)
cv2.waitKey(0)
cv2.destroyAllWindows()
So the complte code
#!/usr/bin/python3
import cv2
import numpy as np
img = cv2.imread("lena.png", cv2.IMREAD_UNCHANGED)
face = np.ones((200, 150, 3))
cv2.imshow("Demo", img)
face = img[200:400, 200:350]
cv2.imshow("face", face)
cv2.waitKey(0)
cv2.destroyAllWindows()
Related links: