In order to extract a meaningful amount of information from the images, we need to make sure our feature extractor extracts features from all the parts of a given image. Consider the following image:

If you extract features using a feature extractor, it will look like this:

If you use Dense detector, it will look like this:

We can control the density as well. Let's make it sparse:

By doing this, we can make sure that every single part in the image is processed. Here is the code to do it:

import cv2
import numpy as np

class DenseDetector(object):
    def __init__(self, step_size=20, feature_scale=40, img_bound=20):
        # Create a dense feature detector
        self.detector = cv2.FeatureDetector_create("Dense")

        # Initialize it with all the required parameters
        self.detector.setInt("initXyStep", step_size)
        self.detector.setInt("initFeatureScale", feature_scale)
        self.detector.setInt("initImgBound", img_bound)

    def detect...