Building Features from Image Data in Microsoft Azure Flashcards
How is the feature descriptor for scale-invariant feature transform (SIFT) structured?
8 bin histogram per keypoint
128 element vector per keypoint
36 bin histogram per keypoint
3,780 element feature vector per image
128 element vector per keypoint
What is a common use case for the Histogram of Ordered Gradients?
Pearson detection
Sentiment analysis
Image stitching
Image classification
Image classification
What term is best described as an individual measurable property or characteristic?
Feature
Activation
Descriptor
Kernel
Feature
If you want to detect the exact same feature in multiple images on an edge device with limited computing power, which approach would work best?
Learning
Non-hybrid
Hybrid
Non-learning
Non-learning
What is a learning-based approach to computer vision?
SURF
Convolutional neural network (CNN)
Object-based Resource Browser (ORB)
Ccale-invariant feature transform (SIFT)
Convolutional neural network (CNN)
What type of (CNN) layer reduces complexity and the risk of overfitting by downsampling the input data?
Flatten layer
Convolutional layer
Dropout layer
Pooling layer
Pooling layer
Which type of activation function should be used for classification tasks when more than two classes are being considered?
sigmoid
relu
tanh
softmax
softmax
You trained a model to 98% accuracy with a large adequate data set. When you validate the data against the test data set, it performs much worse, however. What model deficiency is likely occurring here?
High-loss variance
Underfitting
Sigmoid effect
Overfitting
Overfitting
What is the name for a matrix that is applied to each pixel within an image for the application of a filter or the detection of a feature?
Activation function
Kernel
Gaussian
Pool
Kernel
What term describes a group of images at the same scale but with different sigma values within the scale-invariant feature transform (SIFT) scale-space extrema detection step of SIFT?
Ordered gradients
Bin
Octave
Gaussian
Octave
