Table of Contents
What is semantic segmentation and why is it used for pixel-level classification?
Prepare for the AI-900 exam with a clear explanation of semantic segmentation. Learn how this computer vision task classifies every pixel in an image into specific categories and understand its key differences from image classification, object detection, and instance segmentation in Azure AI.
Question
In Azure AI, which computer vision task involves classifying every pixel in an image into a specific category, such as “road,” “sky,” or “tree,” to generate a segmentation map?
A. Image classification
B. Semantic segmentation
C. Instance segmentation
D. Object detection
Answer
B. Semantic segmentation
Explanation
The computer vision task described is B. Semantic segmentation. This technique involves assigning a class label to every single pixel within an image.
Understanding Semantic Segmentation
Semantic segmentation is a dense prediction task where the goal is to create a pixel-level map of an image. Each pixel is categorized into a class, such as “road,” “sky,” “tree,” or “person.” The output is a “segmentation map,” which is an image of the same size as the input, where each pixel’s color corresponds to the class it belongs to. The key aspect of semantic segmentation is that it understands what is in the image at a pixel level, but it does not differentiate between separate instances of the same object. For example, if there are two cars in an image, all pixels belonging to both cars would simply be labeled as “car.”
Distinguishing from Other Computer Vision Tasks
- Image classification: This is a simpler task that assigns a single label to the entire image. For example, it would label an image as “contains a car” but would not identify the location of the car or the specific pixels that make it up.
- Instance segmentation: This is a more advanced task that goes one step further than semantic segmentation. It also classifies every pixel, but it additionally distinguishes between different instances of the same object. In the two-car example, instance segmentation would label the pixels of the first car as “car 1” and the pixels of the second car as “car 2.”
- Object detection: This task identifies the location of objects within an image by drawing bounding boxes around them and assigning a class label to each box. It answers “what” and “where” but does not provide precise pixel-level outlines of the objects.
Microsoft Azure AI Fundamentals AI-900 certification exam practice question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Microsoft Azure AI Fundamentals AI-900 exam and earn Microsoft Azure AI Fundamentals AI-900 certification.