From: https://github.com/ducha-aiki/affnet/blob/master/examples/SIFT-AffNet-HardNet-kornia-matching.ipynb

import cv2
import numpy as np 

def extract_sift_keypoints_upright(img, n_feat = 5000):
    sift = cv2.xfeatures2d.SIFT_create(2 * n_feat, 
            contrastThreshold=-10000, edgeThreshold=-10000)
    keypoints = sift.detect(img, None)
    response = np.array([kp.response for kp in keypoints])
    respSort = np.argsort(response)[::-1]
    kpts = [cv2.KeyPoint(keypoints[i].pt[0], keypoints[i].pt[1], keypoints[i].size, 0) for i in respSort]
    kpts_unique = []
    for x in kpts:
        if x not in kpts_unique:
            kpts_unique.append(x)
    return kpts_unique[:n_feat]
  
#
def match_snn(desc1, desc2, snn_th = 0.8):
    index_params = dict(algorithm=1, trees=4)
    search_params = dict(checks=128)  # or pass empty dictionary
    matcher = cv2.FlannBasedMatcher(index_params, search_params)
    matches = matcher.knnMatch(desc1.astype(np.float32), desc2.astype(np.float32), k=2)
    good_matches = []
    for m,n in matches:
        if m.distance < snn_th * n.distance:
            good_matches.append(m)
    return good_matches