Python Opencv how to improve SIFT output?

Question

        train_image = 'train_image location'
        sift = cv2.SIFT_create() 
        gray = cv2.cvtColor(train_image, cv2.COLOR_BGR2GRAY)
        (kp, descs) = sift.detectAndCompute(gray, None)
      

I have a dataset of 39,209 traffic sign pictures for my train model. When i try to get SIFT features from them 3131 of the pictures were unable to create descs.After that i tried resize the images that had problems with this code

resized = cv2.resize(train_image,(256,256))

The number went from 3131 down to 2613. What can i else do to make the SIFT better?

Answer 1

              resized = cv2.resize(roi_image,(256,256))
              alpha = 2.5 # Contrast control (1.0-3.0)
              beta = 50 # Brightness control (0-100)
              adjusted = cv2.convertScaleAbs(resized, alpha=alpha, beta=beta)
              sift = cv2.SIFT_create() 
              gray = cv2.cvtColor(adjusted, cv2.COLOR_BGR2GRAY)
              (kp, descs) = sift.detectAndCompute(gray, None)

I found out that changing the brightness and contrast could help.It made the troublesome data go down from 2613 to araound 300.