Lecture 6: Object Recognition & Line Extraction

Question 1

Concerning the Bag of Words(BoW) approach, what are the 2 questions asked?

Answer 1

Is this image in my database?
Robot: Have I been to this place before?

Question 2

Concerning the Bag of Words(BoW) approach, what 3 analogies can be used from text retrieval?

Answer 2

Visual words, Vocabulary of Visual Words, “Bag of Words”(BoW) approach

Question 3

What is the succession order in image retrieval?

Answer 3

Image collection –>Extract features –>Cluster Descriptors

Question 4

What is image retrieval in BoW?

Answer 4

represents a more general problem of object or place recognition

Question 5

What can we do with Image Retrieval?

Answer 5

describe a scene as a collection of words and look up in the database for image with a similar collection of words

Question 6

What if we need to find an object/scene in a database of millions of images?(2 things)

Answer 6

• Build Vocabulary Tree via hierarchical clustering
• Use the Inverted File system: a way of efficient indexing

Question 7

What is the format of the Inverted File System?

Answer 7

Each node in the tree is associated with a list of images containing an instance of this node.

Question 8

Concerning Vocabulary trees, what is K-means clustering? What is minimized?

Answer 8

partitioning a point cloud into k clusters, such that each point belongs to one cluster.
the Sum of Squared Euclidean Distances between points and their nearest cluster-centers

Question 9

What is the algorithm for Vocabulary trees?

Answer 9

• Randomly initialize k cluster centers

Question 10

Until Convergence, what 2 things do you do in the algorithm for Vocabulary?

Answer 10

-Assign each data-point to its nearest cluster-center
-Re-compute each cluster-center as the means of all points assigned to each cluster

Question 11

In Vocabulary tree construction, what does each node represent?

Answer 11

a cluster of descriptors

Question 12

In Vocabulary tree construction, what does each leaf represent?

Answer 12

a visual word

Question 13

What does an inverted file DB list? Inside it, what does each word point to?

Answer 13

all possible visual words.
a list of images where this word occurs.

Question 14

What does a Voting array have?

Answer 14

as many cells as the images in the DB - each word in query image votes for an image

Question 15

what is tf-idf acronym for?

Answer 15

Term
Frequency
-Inverse
Document

Question 16

what does tf-idf measure?

Answer 16

the importance of a visual word inside a document(as part of a document DB)

Question 17

what is term frequency?

Answer 17

frequency of word in image

Question 18

why should tf-idf of word in image j ?

Answer 18

use it to weigh the importance of each word when voting for corresponding image

Question 19

What are the 3 main problems in line extraction from a point cloud?

Answer 19

How many lines are there?
(Which points belong to which line?)Segmentation
(Given points that belong to a line, how to estimate the line parameters?) Line Fitting/Extraction

Question 20

What are 4 algorithms used in line extraction from a point cloud?
(S-a-M, LR, 5, H-T)

Answer 20

Split-and-merge,
Linear regression,
RANSAC,
Hough-Transform

Question 21

When it comes to split-and-merge, how does it work?

Answer 21

Iterative end-point-fit: simply connects the end points for line fitting

Question 22

When it comes to line-regression, how does it work?

Answer 22

“Sliding window” of size N points; fit line-segment to all points in each window; merge overlapping line segments + re-compute line parameters for each segment.

Question 23

What does RANSAC stand for?

Answer 23

RANSAC = RANdom SAmple Consensus

Question 24

What is RANSAC?

Answer 24

A generic & robust fitting algorithm of models in the presence of outliers (i.e., points which do not satisfy a model)

Question 25

Where can RANSAC be applied?

Answer 25

in general to any problems, where the goal is to identify the inliers which satisfy a predefined model

Question 26

What are the 6+ typical RANSAC applications?

Answer 26

line extraction from 2D range data,
plane extraction from 3D data,
feature matching,
structure from motion,
camera calibration,
homograph estimation, etc

Question 27

How does RANSAC work?(2 adjectives)

Answer 27

is iterative and non-deterministic –> the probability to find a set free of outliers increases as more iterations are used.

Question 28

What is the drawback of RANSAC?

Answer 28

a non-deterministic method, results are different between runs

Question 29

What are the 5 steps of RANSAC?
(S.C.C.S.R)

Answer 29

1. Select sample of 2 points at random
2. Calculate model parameters that fit the data in the sample
3. Calculate error function for each data point
4. Select data that supports current hypothesis
5. Repeat sampling
   Set with the maximum number of inliers obtained after k iterations

Question 30

How many iterations does RANSAC need?

Answer 30

k = #
* Ideally: check all possible combinations of 2 points in a dataset of N points.
* Number of all pairwise combinations: ?
* Computationally infeasible if N is too large.
* Example: 10,000 points to fit a line through –> need to check all
10,000* 9,999/2 = 50 million combinations!

Question 31

In RANSAC, do we really need to check all combinations, or can we stop after some iterations?

Answer 31

• Checking a subset of combinations is enough if we have a rough estimate of the percentage of inliers in our dataset.

Question 32

In Hough-Transform, what do points vote for?

Answer 32

plausible line parameters

Question 33

What does Hough-Transform do?

Answer 33

maps image-space into Hough-space

Question 34

What is Hough-space?

Answer 34

voting accumulator, parameterized w.r.t, line characteristics

Question 35

What does a line in image space correspond to in Hough space?

Answer 35

a point

Question 36

What does a point in image space correspond to in Hough space?

Answer 36

a downward line

Question 37

In Hough-Transform, where is the line that contains both X0,Y0 and X1,Y1?

Answer 37

It is the intersection of the lines b = -x0m+ y0 and b = -x1m+ y1