Microsoft Certified Azure AI Fundamentals · Free Practice Question Medium
Question 25
A city is developing a smart surveillance system to monitor public spaces for safety. The system uses a convolutional neural network (CNN) to analyze live video feeds. One key step in processing each frame is applying a small kernel (filter) that “slides” over the image to highlight important features like edges, which are crucial for detecting unusual objects or movements.
In this context, what is the primary purpose of applying the kernel via a convolution operation on each image?
-
A
To increase the overall resolution of the image by adding extra pixels.
-
B
To slide over the image and compute localized weighted sums that emphasize specific features
-
C
To directly assign labels to objects in the image based on predetermined patterns.
-
D
To convert the image from color to grayscale so that only intensity values are used.
Reveal correct answer
Correct answer: B
Explanation
The primary purpose of applying a kernel via a convolution operation on each image is to slide the kernel over localized regions of the image and compute weighted sums of pixel values. This process creates a feature map that emphasizes specific features such as edges, textures, or other patterns that are crucial for object detection and scene understanding in a smart surveillance system.
How It Works:
Sliding the Kernel: The convolution operation involves moving a small kernel (or filter) across the image, one small region at a time.
Localized Computation: For each position, the kernel performs element-wise multiplication with the overlapping pixels, and the results are summed to produce a single output value.
Feature Extraction: This process captures localized features (e.g., edges, corners, textures) which are fundamental for detecting unusual objects or movements in the video feeds.
Why Other Options Are Less Suitable:
Converting an image to grayscale (option A) is a separate process and does not utilize the kernel's capability to extract specific features.
Directly assigning labels to objects (option C) is typically handled by later layers in the CNN, not by the convolution operation itself.
Increasing image resolution (option D) is unrelated to the convolution process, which is meant for feature extraction, not for image upscaling.
Microsoft Documentation References and Further Readings
Azure Cognitive Services – Computer Vision
A. Increasing the overall resolution of the image by adding extra pixels is not the primary purpose of applying a kernel via a convolution operation. The convolution operation is used to compute localized weighted sums that emphasize specific features in the image, such as edges, to improve the detection of important features in the smart surveillance system.
B. The primary purpose of applying a kernel via a convolution operation on each image is to slide over the image and compute localized weighted sums that emphasize specific features. This process helps highlight important features like edges, which are essential for detecting unusual objects or movements in the smart surveillance system.
C. Directly assigning labels to objects in the image based on predetermined patterns is not the primary purpose of applying a kernel via a convolution operation. The convolution operation is used to compute localized weighted sums that emphasize specific features in the image, such as edges, to enhance object detection capabilities.
D. Converting the image from color to grayscale is not the primary purpose of applying a kernel via a convolution operation. The kernel is used to compute localized weighted sums that emphasize specific features in the image, such as edges, which are crucial for object detection in smart surveillance systems.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
