SEGFAULT using SIFT in Linux 5.4.53, Python 3.8.5

[blakev]

System information (version)

OS: Arch Linux x86_64
Kernel: 5.4.53-1-lts
Uptime: 3 hours, 16 mins
Packages: 2145 (pacman)
Shell: bash 5.0.17
Resolution: 2048x1152, 2048x1152, 3840x2160, 2560x1440
DE: Xfce
WM: Xfwm4
WM Theme: Numix-Frost-Light
Theme: Adwaita-dark [GTK2], Numix-Frost-Light [GTK3]
Icons: elementary-xfce-dark [GTK2], Numix-Square [GTK3]
Terminal: alacritty
CPU: AMD Ryzen Threadripper 2950X (24) @ 3.500GHz
GPU: NVIDIA GeForce GTX 1080 Ti
GPU: NVIDIA GeForce GTX 1070 Ti
Memory: 13031MiB / 31972MiB

Python 3.8.5
gcc version 10.1.0 (GCC)
opencv-contrib-python==4.3.0.36
opencv-contrib-python-headless==4.3.0.36

Detailed description

SEGFAULT when trying to use SIFT algorithm, detectAndCompute.

Steps to reproduce

• 
• 

import cv2

a = cv2.imread('/tmp/a.png')
b = cv2.imread('/tmp/b.png')

sift = cv2.xfeatures2d_SIFT()
k_a, d_a = sift.detectAndCompute(a, None)
k_b, d_b = sift.detectAndCompute(b, None)

Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues,
  answers.opencv.org, Stack Overflow, etc and have not found solution
• I updated to latest OpenCV version and the issue is still there
• There is reproducer code and related data files: videos, images, onnx, etc

[asmorkalov]

SIFT algorithm migrated back to core OpenCV repository and you do not need to use contrib since release 4.4. Also the Algorithm is implemented subclass of cv::Algorithm in C++ code and you need to use create() call to initialize it correctly. Proper code:

import cv2

a = cv2.imread('a.png')
b = cv2.imread('b.png')

sift = cv2.SIFT().create()
k_a, d_a = sift.detectAndCompute(a, None)
print(k_a, d_a)
k_b, d_b = sift.detectAndCompute(b, None)
print(k_b, d_b)

[blakev]

@asmorkalov thanks for the advice. Calling create() didn't seem to help, since opencv-python is still targetting 4.3.0 the cv2.SIFT() isn't there.