Quiz
FSIV
Create a free account to finish this quiz
Save your progress and see your final score.
Already have an account? Sign in
Questions in this quiz (59)
1Natural light is the best possible in all cases:
- A-True
- B-Flase
2The observed color of an object depends exclusively on its reflective properties.
- A-True
- B-False
3The smaller the focal distance, the larger the field of vision.
- ATrue
- BFalse
4Select the type of sensors a camera can have.
- ACCD
- BCMOSS
- CChromatic
- DGeometric
5Applying the linear transformation: y+30:
- A-Makes the image brighter.
- B-Makes the image darker
- C-Makes the image larger in all the two dimensions.
- D-Makes the image larger in the horizontal dimension.
6A linear pixel-wise operation is used to:
- Asave the image to a file.
- Bchange the image illuminatio.
- Creduce the image size.
- Dconvert the color space
7Averaging a set of images taken from a static camera can be used to:
- AIncreasing image sharpness
- BBackground substraction.
- CRemove gaussian noise
- DChange the image color space from RGB to Gray
8Interpolation is used to:
- A-Remove Gaussian noise.
- BEqualize the image.
- C-Obtain the value of subpixels.
- D-Enhance the image.
9
A box filter is used to:- A-Remove high frequencies.
- B-Equalize the histogram.
- C-average the values around a pixel.
- D-Find the maximum value of a pixel
10Which are the parameters of a Gaussian filter?::
- A-The size of the filter.
- Bbicubic
- C-The size of the filter and sigma.
- D-bilinear.
- E-Only sigma
11A bilateral filter:
- A- Increase image contrast.
- B- smooth images preserving edges.
- CReduces the image size.
- Dsmooth the image removing color
12What is a colour
- Aa) A human perception
- Bb) A property of the objects
13What is the moment M00
- Aa) The sum of all pixel intensities
- B0
14What is the gradient
- Aa) Low filter to define the structurs
- Bb) To have a measure of the temperature
- Cc) High-filter to define the shape
15Choose the right order of the main steps of SIFT algorithm for keypoint detection:
- ADifference of Gaussians --- Third
Convolution with Gaussians --- Second
Selection of maxima and minima in DOG space --- Fourth
Pyramid of scales --- First - BDifference of Gaussians --- Third
Convolution with Gaussians --- First
Selection of maxima and minima in DOG space --- Fourth
Pyramid of scales --- Second - CDifference of Gaussians --- Third
Convolution with Gaussians --- Fourth
Selection of maxima and minima in DOG space --- Second
Pyramid of scales --- First - DDifference of Gaussians --- Third
Convolution with Gaussians --- Second
Selection of maxima and minima in DOG space --- First
Pyramid of scales --- Fourth
- ADifference of Gaussians --- Third
16The most common algorithm for keypoint descriptor matching is known as Seleccione una:
- Aa. Harris
- Bb. Lucas-Kanade
- Cc. Nearest Neighbour
- Dd. HOG
17Select the computer vision task where the object of interest is finer-grained (more detailed) defined: Seleccione una:
- Aa. Localisation
- Bb. Segmentation
- Cd. Categorization
- De. Classification
18Select the true statement about object categorization methods:
- Aa. HMAX descriptor does not take into account the spatial location of patches in the original image.
- Bb Part-based models are very computational efficient at test time.
- Cc. HMAX descriptor takes into account the spatial location of patches in the original image by using grids.
- Dd. In part-based models the relative location of parts is not so important.
19The term "inter-class" variability refers to...
- Aa. the visual difference between samples of different categories.
- Bb. the visual difference between samples of the same category.
- Cd. the difference in colour of objects in an image.
- De. the difference in scale of objects in an image.
20When using kNN algorithm for classification, the choice of k as an odd number (e.g. 5) guarantees that a draw is not possible in terms of
neighbour class assignment, improving the probability of success of our system.- ATrue
- BFalse
21For comparing normalized histograms, the preferred distance metric is... Seleccione una:
- Aa. Chi-square
- Bb. Hamming
- Cc. Euclidean
- De. Mahalanobis
22BoVW model can be applied not only with SIFT keypoints but also others like SURF. Seleccione una:
- ATrue
- BFalse
23We can say that k-means is a deterministic algorithm, i.e. each time we execute it for a given k the final result does not change.
- ATrue
- BFalse
24Choose the false statement about PHOW:
- Ab. PHOW can be used not only with SIFT descriptors but also with Others like SURF.
- Bc. PHOW representation can be improved by using colour information in the HSV space.
- Cd. The first step in PHOW is to compute a pyramid of images of different resolution.
- De. PHOW encodes some spatial information thanks to the grid.
25For a fixed dictionary size k, select the method that provides a larger (more dimensions) image descriptor: Seleccione una:
- Aa. VLAD
- Bb. Fisher Vector
- Cc. Dense SIFT
- De. BoVW
26Choose the correct statement:
In VLAD method, ...- Aa. the final descriptor does not need to be normalized, due to its nature.
- Bc. the final descriptor is L2-normalized.
- Cd. the final descriptor is Ll-normalized.
- De. the final descriptor is L0-normalized.
27The category Of classifiers where only some of the training samples are labelled is known as Seleccione una:
- Aa. unsupervised
- Bb. semi-supervised
- Cc. It is not possible such situation
- Dd. supervised
28Which of the following dataset partition types is used to tune the hyperparameters of the classifier? Seleccione una:
- Aa. validation
- Bb. training
- Cc. test
- Ddeployment
29As a general rule, in order to be able to localise objects in images at different scales,... Seleccione una:
- Aa. we train several detectors, one per possible discretised scale.
- Bc. we use a cascade of classifiers.
- Cd. we convolve the target image with different Gaussian kernels, one per possible scale.
- De. we apply a single detector at diferent image resolutions.
30Which of the following classifiers can deal with multiclass problems in a natural way (i.e_ without something additional)?
- Aa. AdaBoost
- Bb. Decision trees
- Cd. None Of them
- De. Support Vedor Machines
31Which of the following classifiers does not need to train any parameter? Seleccione una:
- Aa. AdaBoost
- Bb. SVM with linear kernel c. Decision trees
- Cd. 5-NN
32Select the false statement about STIPs:
- Aa. STIPs computation follows the same steps that SIFT keypoint detection.
- Bb. STIPs are equivalent to Harris points but in a 3D space.
- Cc. STIPs correspond to strong changes in space and time.
- De. STIPs computation require the estimation of gradients in three orthogonal directions.
33The homogeneous coordinates is a mathematical method to:
- Aa. obtein the distance in euclidean spaces.
- Bb. compact manifolds
- Cc. measure distances in geodesic spaces such as rotations.
- Dd. to represent linear operations that can be compacted in advance
34A set of transformations such as 3D rotations and translations can be precalculated in a single 4x4 matrix:
- ATrue
- BFalse
35The pin-hole model:
- Aa. represent the camera location in a arbitray reference system.
- Bb. represent the ideal camera with no distorsion.
- Cc. model the rodrigues rotation.
36The intrinsic parameters are:
- Aa. the focal length, optical center and distorsion coefficients.
- Bb. the rotation and translation components.
- Cc. none of the above
- Dd. the focal lenght and optical center.
37The intrinsic parameters.
- Aa. change whenever the camera moves.
- Bb. never changes.
- Cd. changes only if we replace the lens.
38The extrinsic parameters:
- Aa. never change.
- Bb. change whenever illumination changes.
- Cc. represent the camera location with respect to an arbitrary reference system.
- Dd. represent the camera interval of confidence.
39Calibrating a camera is a process to
- Aa. obtain the camera translation
- Bb. obtain the camera response to illumination changes
- Cc. obtain the camera rotation
- Dd. obtain the camera intrinsic parameters
40Perspective-N-Point is the problem of
- Aa. Estimating the camera intrinsic parameters
- Bb. calibrate the camera
- Cc. rotating the camera around its axis
- Dd. finding the camera extrinsics from a set of know 3D points and its 2D projections
41The Rodrigues representation
- Aa. allows to express rotations as a 3 component vector
- Bb. is a measure of the calibration error
- Cc. determines the full camera location in the space
- Dd. is a 4x4 matrix that can transform a 3D point between reference systems
42Given the projection of a point in a camera, its projection in another camera Seleccione una:
- Aa. can be determined even if the 3d position of the point is unknown
- Bb. can be determined by the second camera instrinsics
- Cc. Is inversely proportional to the distance of the epipole
- Dd. lies along a line called epipolar line
43In stereo vision, matches are seek in the right image by Seleccione una:
- Aa. using a HOG descriptor
- Bb. using a epipolar representation of the right image
- Cc. comparing small window patches along the epipolar line
- Dd. detecting keypoints
44In order to estimate the 3d location of a point
- Aa. it is required to know its projections in at least one images and the camera's intrinsic parameters
- Bb. it is required to know its projections in at least three images and the camera's extrinsic parameters
- Cc. it is required to know its projections in at least two images and the camera's intrinsic parameters
- Dd. it is required to know its projections in at least two images and the camera's intrinsic and extrinsic parameters
45In stereo vision systems, the system is properly calibrated in order to ensure that:
- Aa) The epipolar lines of both images are perpendicular
- Bb) Points are parallel to the optical center
- Cc) Both cameras have the same extrinsics
- Dd) The epipolar lines of both images are parallel
46The Kinect 3D camera is a type of structured light system:
- ATrue
- BFalse
47The Solve Perspective from n points:
- Aa) Finds the 3D position of a set of points from multiple views
- Bb) Is a method that finds the camera intrinsics given a two points in the 3D space
- Cc) Is a method that finds the camera intrinsics given a set of 3D-2D correspondences
- Dd) Is a method that finds the camera extrinsics given a set of 3D-2D correspondences
48SLAM is the acronym of Simulated Localization and Making:
- ATrue
- BFalse
49The Structure From Motion problema consists in:
- Aa) Jointly estimating the camera extrinsics and the 3D reconstruction from a set of views taken from unknown locations
- Bb) None of the previous ones
- Cc) Calibrating a camera using a chessboard panel
- Dd) Estimating the camera intrinsics and the 3D reconstruction from a set of views taken from known locations
50The HSV color space:
- Aa. Is a cubic space tha allows a faster analysis of the image
- Bb. All answers are incorrect
- Cc. Is the preferred method for low-pass filters
- Dd. Splits the information into two chromatic channels and one intensity channel.
51The magnitude of the image derivative, seleccione una:
- Aa. has a low response in bright image regions
- Bc. has a high response in regions of the image that represent transitions
- Cd. has a high response in dark image regions
- De. has a high response in regions of the image that have constant values
52A median filter: seleccione una:
- Aa. Is a global pixel-wise operation
- Bb. Is a non lineal filter
53A median filter, Seleccione una:
- Ab. Increase the global image brightness
- Bc. remove noise by selecting the central value of the sorted pixels
- Cd. add noise by averaging the values around ecah pixel
- De. add noise by selecting the cenral value of the sorted pixel
54Wich of this filters removes salt and peper noise? seleccione una:
- Ab. binomial filter
- Bc gaussian filter
- Cd. bilateral filter
- De. a median filter
- Ef. a box filter
55the sobel filter, seleccione una:
- Aa. computes the magnitude of the derivate of a pixel
- Bb. computes the normalized brightness around a pixel
- Cc. computes the partial derivates of a pixel
- De. computes the squared differences between two images
56Wich is the best filter to remove gaussian noise?
- Aa. Gaussian filter
- BOtro
57We can assert about image gradient that ...
- Aa. can be computed only by using a LoG filter.
- Bb. is a low-level filter that provides information about structures in the image
- Cc. is a high level feature used to detect objects in images
- Dd. can be computed only on RGB images.
58The morphological dilation is the process of, seleccione una:
- Aa. unsharp images using a structuring element mask
- Bb. computing the minimum in the neighborhood of a pixel using a structuring element mask
- Cc. computing the maximum in the neighborhood of a pixel using a structuring element mask
- Dd. computing the average in the neighborhood of a pixel
59Closing is
- Aa. dilation + erosion
- Bb. erosion + dilation
- Cc. erodion + erosion
- Dd. dilation + dilation
Information
- Description
- Ingeniería Quiz on FSIV, created by Pablo Luque on 02/01/2019.
- Author
- Pablo Luque
- Language
- Español
- Views
- 73
- Tags
- Rating
- No ratings yet Rate this quiz