fkl;sdj Flashcards by Joe D.

Computer vision is concerned with modeling and replicating human vision using
computer software and hardware?
a. Can not say
b. FALSE
c. TRUE
d. Can be true or false

How well did you know this?

Not at all

Perfectly

. Computer vision is a discipline that studies how to reconstruct, interrupt and understand a
3d scene from its ________.?
a. 1d images
b. 3d images
c. 4d images
d. 2d images

How well did you know this?

Not at all

Perfectly

The input and output of image processing are?
a. depends on input
b. image only
c. signal and image
d. signal only

How well did you know this?

Not at all

Perfectly

Which of the following studies various techniques to classify patterns?
a. Pattern Recognition
b. Image Recognition
c. Photogrammetry
d. Image Processing

How well did you know this?

Not at all

Perfectly

Which of the following is an Applications of Computer Vision?
a. Robotics
b. Medicine
c. Security
d. All of the above

How well did you know this?

Not at all

Perfectly

the spatial coordinates of a digital image(x,y) are proportional to—-?
a. Noise
b. Brightness
c. contract
d. Position

How well did you know this?

Not at all

Perfectly

Among the following image processing techniques which is fast, precise and flexible?
a. Optical
b. Photographic
c. Digital
d. Electronic

How well did you know this?

Not at all

Perfectly

What is pixel?
a. Pixel is the elements of an analog image
b. Pixel is the elements of a digital image
c. Pixel is the cluster of a digital image
d. Pixel is the cluster of an analog image

How well did you know this?

Not at all

Perfectly

In computer graphics, given the 3D scene composed of objects, materials, certain
semantic pose and motion we are concerned with rendering a photorealistic 2D image of
that scene.?
a. True
b. false

How well did you know this?

Not at all

Perfectly

computer vision is the invers of computer graphics?
a. false
b. True

How well did you know this?

Not at all

Perfectly

image processing can do what computer vision do?
a. false
b. True

How well did you know this?

Not at all

Perfectly

challenges of CV are—-?
a. Viewport
b. illumination
c. A,B
d. B,C
e. occlusion

How well did you know this?

Not at all

Perfectly

——–image data doesn’t matter as long as there are enough features remaining in
image
a. Refracted
b. missing
c. transmitted
d. reflected

How well did you know this?

Not at all

Perfectly

——–can be used to initialize fitting for more complex models?
a. Affine transformation
b. 2D transformation
c. Fourier transformation
d. 3D transformation

How well did you know this?

Not at all

Perfectly

we need ———geometry because recovery of structure from one image is inherently
ambiguous?
a. single view
b. geometrical correct view
c. line-view
d. multi-view

How well did you know this?

Not at all

Perfectly

if we can scale the entire scene by a factor k and at the same time scale the camera
matrices by the factor 1/k, the projections of the scene point in the image—?
a. remain unchanged
b. convert to 3D
c. convert to 2D
d. change randomly

How well did you know this?

Not at all

Perfectly

rays passing through the center of the lens are ——?
a. deviated
b. not deviated
c. scattered
d. reflected

How well did you know this?

Not at all

Perfectly

translation transformation matrix can affect ——-image ?
a. scale
b. position
c. dimension
d. all of the above

How well did you know this?

Not at all

Perfectly

scaling matrix keeps the relative distance of image…….?
a. all above
b. decrease if less than one
c. constant for both coordinate
d. the same
e. increase if greater than 1

How well did you know this?

Not at all

Perfectly

All ……. Rays converge to one point on a plane located at the focal length f
a. Parallel
b. Diverged
c. Perpendicular
d. Intersect

How well did you know this?

Not at all

Perfectly

……. Projection occurs when camera at far point
a. Orthogonal
b. Week perspective
c. Perspective
d. A,B

How well did you know this?

Not at all

Perfectly

———-projection camera coordinate and image coordinate coincide?
a. perspective
b. A, B
c. week perspective
d. orthogonal

How well did you know this?

Not at all

Perfectly

Which of the following is the role played by segmentation in image processing?
a. Deals with techniques for reducing the storage required saving an image, or the
bandwidth required transmitting it
b. Deals with partitioning an image into its constituent parts or objects
c. Deals with extracting attributes that result in some quantitative information of interest
d. Deals with property in which images are subdivided successively into smaller regions

How well did you know this?

Not at all

Perfectly

Region of Interest (ROI) operations is generally known as _______?
a. Dilation
b. Shading correction
c. Masking
d. None of the Mentioned

How well did you know this?

Not at all

Perfectly

f we apply transformation to an object matrix in the following order R.S.T then to return to the original object use---? a. inv(T)inv(S)inv(R) b. Inv(T)inv(R)inv(S) c. A and B d. inv(R).inv(S).Inv(T)

The degree of freedom is proportional to---? a. All of the above b. homogeneous coordinate c. type of transformation d. object coordinate

To apply translation to any object you need to----? a. multiply the object matrix with translation matrix b. Add translation matrix to object matrix c. draw the object d. move the object to the origin

Which of the following is the abbreviation of JPEG? a. Joint Photographic Expansion Group b. Joint Photographic Expanded Group c. Joint Photographs Expansion Group d. Joint Photographic Experts Group

To apply rotation to any object you need to----? a. multiply the object matrix with rotation matrix b. Add rotation matrix to object matrix c. draw the object d. move the object to the origin

We use homogeneous coordinate for any type of transformation a. False b. True

f the object matrix is A and required to rotate (R), and scale (S) object then the operation is-? a. R.S.X b. Non c. T-1.S.R.T.X d. S.R.X

Which are recognized by vision a. Motion b. B and C c. Objects d. Activities

Adding an imaginary plane to the object coordinate is for ---? a. treat all transformation in consistence way b. Non c. adding linearity to transformation d. increasing the dimensions of object coordinate

YOLO algorithm is important because of the following reasons:? a. speed b. accuracy c. Learning capabilities d. all of the above

It refers to making the image less clear or distinct. It is done with the help of various low pass filter kernels. a. Erosion and Dilation of images b. Eroding an image c. Grayscaling of image d. Blurring an image

aims at assigning an image to one of a number of different categories (e.g. car, dog, cat, human, etc.), essentially answering the question “What is in this picture?”. One image has only one category assigned to it.? a. object localization b. recognition c. image classification d. Object detection

is a computer vision task that involves identifying and locating objects in images or videos.? a. image processing b. tracking c. object detection d. recognition

provides the tools for doing just that – finding all the objects in an image and drawing the so-called bounding boxes around them.? a. object localization b. Object detection c. recognition d. image classification

Cropping: Now, suppose I want to extract a second rectangular region from my image starting at x=1,y=48 and ending at x=80,y=69. What is the correct line of code to perform this cropping? a. Crop = image[1:48, 80:69] b. Crop = image[48:69, 1:80] c. Crop = image[80:69, 1:48] d. Crop = image[48:80, 48:69]

How many types of recognition are there in artificial intelligence? a. 2 b. 3 c. 1 d. 4

is a popular metric to measure localization accuracy and calculate localization errors in object detection models.? a. Intersection over Union b. single-shot object detection c. Average Precision d. Two-shot object detection

The RGB tuple (255, 0, 0) codes for red. But OpenCV would actually interpret this color as:? a. Green b. Blue c. orange d. red

Suppose our image has a width of 700 pixels, a height of 550 pixels, and 3 channels, one for each Red, Green, and Blue component. How would we express this image as a NumPy array shape? a. (3,500,700) b. (3, 700, 550) c. (550, 700, 3) d. (700, 550, 3)

How the distance between two shapes can be defined? a. Weighted sum of the shape b. Size of the shape c. None of the mentioned d. Shape context

OpenCV stores RGB pixels in what order? a. RGB b. GRB c. BRG d. BGR

1-The process of Canny edge detection algorithm can be broken down to five different steps which one isn’t from these steps? a. Apply Gaussian filter to smooth the image in order to remove the noise b. Apply double threshold to determine potential edges c. Apply gradient magnitude thresholding or lower bound cut-off suppression to get rid of spurious response to edge detection d. Apply color converting to gray level

The filter that gives more edges details with the same initialization is----? a. Sobel b. Prewitt c. Robert d. Canny

The difference between (Prewitt &Robert &Sobel and Canny) can be---? a. convolution mask b. Angle of Orientation c. steps of algorithm d. Gaussian parameters

a,b

When taken the gradient of the image in the two dimension (X,Y) the edges and the corners divers in-? a. lines of the same color. b. lines with different color. c. same Distribution of Image Gradients. d. Emax and Emin

b,d

The gradient distribution of the edge fil in ---? a. circle with small radius b. circle with Large radius c. Ellipse with small varying of two radius is dimension d. Ellipse with Large varying of two radius dimension

The gradient distribution of the corner fil in ---? a. Ellipse with Large varying of two radius dimension b. Ellipse with small varying of two radius dimension c. circle with small radius d. circle with Large radius

Which filter is used to recognize the image? a. Harries b. candy c. SIFT d. all of the above

Non-maximal suppression is used in --? a. Harries b. candy c. SIFT d. all of the above

The slide of window of size k over the image in Harries detector is known as-----in the build in function? a. Ksize. b. K. c. blockSize. d. all of the above

For image alignment and 2D object recognition you need to-----? a. resize, scale and change the brightness of the image. b. none of the above. c. rotate, scale and change lighting of the image. d. change the orientation and size of the image using interest point

The interest point is-----? a. has rich image content. b. different signature. c. insensitive to positions and lighting d. all of the above

Edges are significant local changes in the image and are important features for analyzing images.? a. True. b. False. c. and also Interesting points. d. all of the above.

Edges typically occur on the boundary between two different regions in an image? a. True. b. False

Are Lines and Edges Interesting point? a. True. b. False.

Are Blobs Interesting point? a. True. b. False

In candy filter as the sigma of the gaussian filter decrease the -------? a. corners we detect. b. matching point we find. c. edges we draw. d. all of the above.

Before determining the matching point between two images using detector function we need to----? a. resize the two images. b. find the interesting point. c. convert the image to gray color. d. all of the above.

For an object detector using sliding window for detecting 5 classes the size of Y matrix will be----? a. 8. b. 10. c. 15. d. none

Sliding window algorithm for object detection and classification we-----? a. pass the grid cell through image repeatedly with different grid size. b. the image part of grid cell is convoluted with predicted output. c. computationally expensive. d. all of the above.

From the real time detection algorithm with high accuracy is---? a. sliding window. b. SIFT. c. you only look once. d. All of the above.

If the detected object lays in more than one boundary box we use--? a. YOLO. b. IOU. c. Sliding window. d. all of the above.

If we detect more than one object in the image we use-----? a. same size box but different color. b. different centers for each box and different color. c. different size box and different color. d. none

the output matrix of the convolution network equal multiple of--------? a. detected box. b. detected object. c. detected landmarks d. all of the above

Assign a single class label to image is--? a. Semantic segmentation. b. Object Detection. c. image classification. d. instance Segmentation

Assign a semantic label to every pixel in the image.? a. Semantic segmentation. b. Object Detection. c. image classification. d. instance Segmentation.

Localize and classify all objects in the image? a. Semantic segmentation. b. Object Detection. c. image classification. d. instance Segmentation

is it easy to detect an object using edges and corners? a. True. b. false. c. hand-engineering make it possible. d. hard.

Unsupervised is used for----? a. image detection. b. object classification. c. YOLO d. none

--------- enables a machine to interact with an environment? a. supervised Learning. b. unsupervised Learning. c. Semi supervised Learning. d. Reinforcement Learning.

------- learning combines techniques? a. supervised Learning. b. unsupervised Learning. c. Semi supervised Learning. d. Reinforcement Learning.

-------enables a machine to explore a set of data. After the initial exploration, the machine tries to identify hidden patterns that connect different variables? a. supervised Learning. b. unsupervised Learning. c. Semi supervised Learning. d. Reinforcement Learning

transfer learning is---? a. movement of learning network to another. b. use Pretrain network for your own dataset. c. fine tune of pretrained network only last layers on your own data. d. retrain a pretrained network on your own data set.

Various kinds of recognition -----? a. instance segmentation. b. pose estimation. c. panoptic segmentation. d. All of the above.

Nearest neighbor classifier drawback of image classification is partially solved by Bag of words? a. True. b. False.

The architecture of a -----is analogous to that of the connectivity pattern of Neurons in the Human Brain and was inspired by the organization of the Visual Cortex.? a. ConvNet. b. neural network. c. both ConvNet,and neural network. d. all of the above.

activation function computes-----? a. the weight of the nodes. b. the output of the node. c. computes the a value. d. the base value.

if the input image size is32X32 and the filter size 4X4 and striding 2 ? a. 28 b. 14 c. 15 d. 32

artificially increase the size of the training set-create a batch of "new" data from existing data by means of translation, flipping, noise--? a. Data augmentation. b. Regularization c. Dropout d. Unsupervised Pre-training

also a regularization method. But different from the above, it is achieved by randomly setting the output of some neurons to zero--? a. Data augmentation. b. Regularization c. Dropout d. Unsupervised Pre-training

use Auto-Encoder or RBM's convolution(restricted Boltzmann machine) form to do unsupervised pre-training layer by layer, and finally add a classification layer to do supervised Fine-Tuning-? a. Data augmentation. b. Regularization c. Dropout d. Unsupervised Pre-training

The relatively small amount of data will cause the model to overfit, making the training error small and the test error particularly large. By adding a regular term after the Loss Function , the overfitting can be suppressed-? a. Data augmentation. b. Regularization c. Dropout d. Unsupervised Pre-training

The disadvantage of---------is that a need is introduced Manually adjusted hyper- parameter? a. Data augmentation. b. Regularization c. Dropout d. Unsupervised Pre-training

------ happens when a model learns the detail and noise in the training data to the extent that it negatively impacts the performance of the model on new data? a. underfitting. b. overfitting. c. optimal fitting. d. none.

------ refers to a model that can neither model the training data nor generalize to new data? a. underfitting. b. overfitting. c. optimal fitting. d. none.

Once the computation for gradients of the cost function w.r.t each parameter (weights and biases) in the neural network is done, the algorithm takes a gradient descent step towards the minimum to update the value of each parameter in the network using these gradients? a. the forward propagation operation. b. The backpropagation operation. c. cost function and regularization. d. all of the above.

---------first network introduced the concept of increasing the number of layers to improve accuracy? a. VGG16 b. ResNet c. InceptionNets d. EfficientNet

increasing the number of layers above 20 could-----? a. cause overfitting. b. prevent the model from converging. c. enables the model to converging d. all of the above.

skip connections is introduced by------? a. VGG16 b. ResNet c. InceptionNets d. EfficientNet

skip connection overcome ----? a. vanishing weights. b. training very deep neural network c. Exploding weights. d. All of the above.

there is an exponential growth in the model parameters--? a. vanishing weights. b. Exploding weights

the parameters of higher layers change significantly while lower layers remain unchanged--? a. vanishing weights. b. Exploding weights.

the model weights may become Nan--? a. vanishing weights. b. Exploding weights.

The model weights may become 0---? a. vanishing weights. b. Exploding weights.

the model experiences avalanche learning.--? a. vanishing weights. b. Exploding weights.

the model learns very slowly--? a. vanishing weights. b. Exploding weights

this network also consists of two auxillary classification outputs? a. VGG16 b. ResNet c. InceptionNets d. EfficientNet

key idea of inception module is to ----? a. increasing number of layers. b. design good local network. c. training with massive data and supercomputer. d. All of the above.

the auxiliary classification outputs used to---? a. detect different types of features b. All of the above. c. inject gradients at lower layers d. recover from overfitting

------scaling method uniformly scales network width, depth, and resolution with a set of fixed scaling coefficients? a. VGG16 b. ResNet c. InceptionNets d. EfficientNet

Neural Style Transfer (NST) uses---? a. convolution network b. ResNet c. transfer learning d. EfficientNet

The activation of a particular layer use----? a. sess.run(model[“conv4_2”]. b. model[“input”].assign (image) c. both are true. d. non

The output image from the Neural Style Transfer depends on the----? a. the included activation layers b. the ratio of content cost added to the ratio of style cost c. the generated cost function d. all of the above

The initial generated image of Neural Style Transfer is----? a. style image b. random generated image c. content image d. none

------ compare how similar Vi is to Vj, if they are highly similar, the expected dot product will be large? a. similar matrix b. Gram matrix c. Style matrix d. all