diff --git a/A2/images/astronauts/Catherine-Coleman.jpg b/A2/images/astronauts/Catherine-Coleman.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..53ad70bc1e9329123ec71503503c92d7ce7dfba1
Binary files /dev/null and b/A2/images/astronauts/Catherine-Coleman.jpg differ
diff --git a/A2/images/astronauts/Eileen-Collins.jpg b/A2/images/astronauts/Eileen-Collins.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..e42f76544783e7746f797f56bacc1ae86028972e
Binary files /dev/null and b/A2/images/astronauts/Eileen-Collins.jpg differ
diff --git a/A2/images/astronauts/Jeffrey-Shears-Ashby.jpg b/A2/images/astronauts/Jeffrey-Shears-Ashby.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..568629ec99fb9a6feda424e7485081f820dc917b
Binary files /dev/null and b/A2/images/astronauts/Jeffrey-Shears-Ashby.jpg differ
diff --git a/A2/images/astronauts/Michel-Tognini.jpg b/A2/images/astronauts/Michel-Tognini.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..19a5ff92d43bdabac1ae12173af51020928f6bb5
Binary files /dev/null and b/A2/images/astronauts/Michel-Tognini.jpg differ
diff --git a/A2/images/astronauts/Steven-Alan-Hawley.jpg b/A2/images/astronauts/Steven-Alan-Hawley.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..cb49f36329d4367147c7ffc0eb03552aebd6f203
Binary files /dev/null and b/A2/images/astronauts/Steven-Alan-Hawley.jpg differ
diff --git a/A2/images/chalsea/Kadse.jpeg b/A2/images/chalsea/Kadse.jpeg
deleted file mode 100644
index aa6964c4b7a6dc0dd0629fab3a768a292bee8cf8..0000000000000000000000000000000000000000
Binary files a/A2/images/chalsea/Kadse.jpeg and /dev/null differ
diff --git a/A2/images/martian/martian.jpg b/A2/images/martian/martian.jpg
deleted file mode 100644
index fc71f5f2f030729f05583c8a7ca09d8d0f1e980f..0000000000000000000000000000000000000000
Binary files a/A2/images/martian/martian.jpg and /dev/null differ
diff --git a/A2/images/result.jpeg b/A2/images/result.jpeg
index 67fb23912e1d0925557aead9ba9c26d6a8c1324b..6e10d4ddc3acc8ad131e68348d5dfe39538576e0 100644
Binary files a/A2/images/result.jpeg and b/A2/images/result.jpeg differ
diff --git a/A2/orb.py b/A2/orb.py
index 3be8dc782983efb1f8e6ba65ffca5d6b2659e9c4..08e63763a9eb30f10bf651f2c053ed1c96e6b679 100644
--- a/A2/orb.py
+++ b/A2/orb.py
@@ -1,9 +1,8 @@
-import math
+import os
from tokenize import String
-from typing import List
+from typing import List, Dict
import cv2 as cv
-import imageio
import matplotlib.pyplot as plt
import numpy as np
@@ -22,6 +21,22 @@ class ORB(object):
self.edit_path = edit_path
self.k = k
+ def load_images(self, dir: String) -> List:
+ """
+ This method concats all images in a dir
+
+ :param dir: The directory to search with all images
+ :return: list of all images in a directory
+ """
+ images = []
+ files = os.listdir(dir)
+ for file in files:
+ img = os.path.join(dir, file)
+ img = cv.imread(img, 0)
+ if img is not None:
+ images.append(img)
+ return images
+
@staticmethod
def euclidean(vector1: List, vector2: List) -> float:
"""
@@ -31,9 +46,7 @@ class ORB(object):
:param vector2: Vector as list
:return: Euclidean distance between vector1 and vector2.
"""
- dist = [(a - b) ** 2 for a, b in zip(vector1, vector2)]
- dist = math.sqrt(sum(dist))
- return dist
+ return np.linalg.norm(np.subtract(vector1, vector2))
def pairwise_arg_min(self, X: List, Y: List) -> np.ndarray:
"""
@@ -46,49 +59,78 @@ class ORB(object):
return np.asarray([np.argmin([self.euclidean(x, y) for y in Y]) for x in X])
def find_clusters(self, X, n_clusters, rseed=2):
- # 1. Randomly choose clusters
+ """
+ This method finds all clusters.
+
+ :param X: Data to be clustered
+ :param n_clusters: amount of clusters
+ :param rseed: random seed
+ :return: All labels for clustering and the centroids.
+ """
rng = np.random.RandomState(rseed)
i = rng.permutation(X.shape[0])[:n_clusters]
centers = X[i]
while True:
- # 2a. Assign labels based on closest center
labels = self.pairwise_arg_min(X, centers)
+ new_centers = np.array([X[labels == i].mean(0) for i in range(n_clusters)])
- # 2b. Find new centers from means of points
- new_centers = np.array([X[labels == i].mean(0)
- for i in range(n_clusters)])
-
- # 2c. Check for convergence
if np.all(centers == new_centers):
break
centers = new_centers
return centers, labels
+ def similar_features(self, descriptors: List, centroids: List, labels: np.ndarray) -> List:
+ """
+ This calculates the similar features between the groundtruth and a given image.
+
+ :param descriptors: Descriptors of the image
+ :param centroids: Centroids of the groundtruth
+ :param labels: Labels of the clusters in the groundtruth
+ :return: List of similar features by count
+ """
+ labels = list(set(labels))
+ zeros = [0] * len(labels)
+ cluster_counts = dict(zip(labels, zeros))
+
+ for des in descriptors:
+ min_index = self.pairwise_arg_min([des], centroids)[0]
+ cluster_counts[min_index] = cluster_counts[min_index] + 1
+
+ return self.dict_to_list(cluster_counts)
+
+ @staticmethod
+ def dict_to_list(target: Dict) -> List:
+ """
+ Turns a dict in a list.
+
+ :param target: Dict to be turned into a list.
+ :return: A list containing all values of a dict
+ """
+ return [value for _, value in target.items()]
+
def get_keypoints(self) -> None:
"""
This method does K-Means with the ORB Keypoints.
:return: None
"""
- img = imageio.imread(uri=self.image_path)
- plt.imshow(img)
-
- # Initiate ORB detector
- orb = cv.ORB_create(nfeatures=1000, scoreType=cv.ORB_FAST_SCORE)
-
- # find the keypoints with ORB
- kp = orb.detect(img, None)
- # compute the descriptors with ORB
- kp, des = orb.compute(img, kp)
- key_points = [k.pt for k in kp]
+ images = self.load_images("./images/astronauts")
+ orb = cv.ORB_create()
+ descriptors = None
+ for img in images:
+ kp, des = orb.detectAndCompute(img, None)
+ if descriptors is None:
+ descriptors = des
+ else:
+ descriptors = np.concatenate((descriptors, des), axis=0)
- X = np.array([list(x) for x in key_points])
+ img = cv.imread(self.image_path, 0)
- centers, labels = self.find_clusters(X, self.k)
- plt.scatter(X[:, 0], X[:, 1], marker='.', s=10, c=labels, cmap='viridis')
+ kp, des = orb.detectAndCompute(img, None)
+ centroids, labels = self.find_clusters(descriptors, self.k)
+ hist_with_bins = self.similar_features(des, centroids, labels)
- plt.scatter(centers[:, 0], centers[:, 1], marker='+', color='red')
- plt.axis('off')
- plt.savefig(self.edit_path, dpi=300, bbox_inches='tight', pad_inches=0)
+ plt.bar(range(len(hist_with_bins)), np.multiply(1 / sum(hist_with_bins), hist_with_bins))
+ plt.savefig(self.edit_path, dpi=300, bbox_inches='tight')
plt.close()