Unit 3: Making Machines See (Computer Vision)

Correct answer: A. Understanding and analysing visual information such as images and videos

Computer vision is the area of artificial intelligence that enables machines to analyse images, videos, and other visual data. It can support tasks such as object, face, and pattern recognition, so option A is correct. Written language is mainly associated with NLP, while audio processing and robotic control are different areas.

Correct answer: A. Classifies the image into a predefined category

In image classification, a model analyses an image and assigns it one or more predefined class labels, such as “cat” or “dog.” Therefore, option A is correct. Drawing boxes describes object detection, while marking pixel-level regions describes image segmentation. Converting an image to audio is unrelated to classification.

Correct answer: A. Identifying objects and their locations in an image

Object detection identifies objects in an image or video and estimates where they are located, commonly by drawing bounding boxes around them. Thus, option A is correct. Image-colour adjustment, sound-volume control, and cleaning keyboard keys are unrelated to the purpose of object detection. Unlike classification alone, detection provides location information as well.

Correct answer: A. The smallest picture element of an image

The word pixel comes from “picture element.” It is a small addressable unit in a digital image, commonly storing a colour or brightness value at a particular position. A rectangular image is formed by arranging many pixels in rows and columns; the number of pixels affects its spatial resolution. A book, mouse button, and sound wave are not visual elements of a digital image.

Correct answer: A. Identifying a face in a photograph

Computer vision enables computers to process and interpret visual information from images or videos. Identifying a face in a photograph is therefore a computer-vision task. Translation and speech understanding mainly involve language or speech processing, whereas calculating an average is a numerical operation.

Correct answer: A. To bring images to a common size

Resizing changes an image’s width and height to specified dimensions. It allows a computer vision model to receive inputs with consistent dimensions, which simplifies processing and comparison. Resizing may affect visual detail, but its main purpose is not to erase all image information.

Correct answer: A. When color information is unnecessary and brightness or intensity is sufficient

A grayscale image represents each pixel using a brightness or intensity value instead of separate color channels. It is suitable for tasks such as edge, shape, or pattern detection when color is not needed. Color information is necessary when objects must be distinguished by their colors.

Correct answer: A. To understand an object's boundary and shape

Edge detection identifies locations with sharp changes in intensity or colour. These locations often indicate object boundaries, helping a computer analyse an object's shape and structure. Erasing pixels, generating sound, and translating text are different tasks, not the purpose of edge detection.

Correct answer: A. Because of risks to privacy, consent, and misidentification

Face recognition uses sensitive biometric information. Collecting or using it without appropriate consent can threaten privacy, while inaccurate or biased results can cause unfair treatment. Human oversight, fairness, privacy protection, and the risk of misidentification therefore require careful ethical consideration.

Correct answer: A. Assisting in identifying possible bone fractures in X-ray images

Computer vision analyzes visual images, so it can assist trained healthcare professionals in identifying possible signs of bone fractures in X-rays. Summarizing notes is mainly a language task, calculating bills is an administrative task, and changing dosage without authorization is unsafe.

Correct answer: A. In crop disease detection and irrigation advice

Artificial intelligence can analyse crop images along with weather and soil data to detect diseases and estimate water requirements. It can therefore support timely treatment and efficient irrigation. The other options describe harmful or undesirable activities, not useful agricultural applications of AI.

Correct answer: A. Computer vision

Computer vision enables an AI system to analyse pixels, shapes, colours, and other visual patterns in images. Detecting spots, discolouration, or other signs on leaves is therefore a computer-vision task. Natural language processing and speech recognition work mainly with language or sound, not leaf images.

Correct answer: A. Identifying pedestrians, vehicles, and road signs using data from cameras and other sensors

AI in an autonomous vehicle processes data from cameras, radar, and other sensors to recognise pedestrians, vehicles, signs, and road conditions. This understanding supports safe navigation and decision-making. The other options describe ordinary mechanical operation or impossible and unsafe claims about autonomous driving.

Correct answer: A. Computer vision

The system is analysing visual information in X-ray images, so its primary AI field is computer vision, specifically medical image analysis. Natural language processing handles language, robotics controls physical machines, and an expert system is a broader rule-based system that is not defined by image input.

Correct answer: A. So the model can recognize objects in different situations

Training images that vary in lighting, size, angle, background, and object position expose the model to realistic conditions. This can improve generalization, allowing it to recognize objects in new images rather than merely memorizing the training examples.

Correct answer: A. To create varied versions of images for training

Image augmentation applies controlled changes such as rotation, flipping, cropping, scaling, or brightness adjustment to existing training images. These variations increase data diversity and can help a computer-vision model generalise to new viewpoints and conditions. Thus, option A is correct; the other options do not describe data preparation or model improvement.

Correct answer: A. Object detection identifies objects and their locations, whereas image classification assigns one or more categories to the whole image.

Object detection identifies individual objects and estimates where they are located, commonly with bounding boxes. Image classification generally assigns one or more labels to the image as a whole. Pixel-level labelling is associated with image segmentation, not ordinary object detection.

Correct answer: A. To show an object’s location and approximate extent

In object detection, a bounding box is usually a rectangle placed around a detected object. Its coordinates show where the object appears and indicate its approximate width and height within the image. Therefore, option A is correct. The other choices concern audio, table editing, or file management and have no role in locating visual objects.

Correct answer: A. To review important decisions and correct errors or bias

Facial-recognition systems can make incorrect identifications or produce biased results. Human oversight allows important decisions to be reviewed, challenged, corrected, or stopped when necessary. Without review, an automated error could unfairly affect an innocent person.

Correct answer: B. Computer Vision

Computer Vision enables AI to identify patterns and abnormalities in images such as X-rays, MRI scans, and CT scans. Natural Language Processing handles text or language, recommendation systems suggest items, and speech recognition processes audio; therefore, Computer Vision is correct.

Correct answer: A. Computer Vision

The system receives X-ray images, extracts visual patterns, and identifies a possible fracture region or category. Analysing and interpreting image content is the governing idea of computer vision, so option A is correct. Natural language processing handles text and language, speech recognition handles spoken audio, and recommendation systems suggest items rather than interpret X-ray structure.

Correct answer: A. Crop disease detection in precision agriculture

The app analyses leaf images, detects visual symptoms such as spots or discoloration, and identifies affected crop areas. This is crop-disease detection used in precision agriculture, so option A is correct. Face recognition also analyses images but concerns people; machine translation works between languages, and music recommendation predicts user preferences rather than plant disease.

Correct answer: A. To make images suitable for the model

Image preprocessing can include resizing an image, removing noise, and normalizing pixel values. Its purpose is to make raw images more consistent and suitable for analysis by a model. In contrast, deleting images or generating sound from them is not an image-preprocessing task.

Correct answer: A. So the model can learn which class each image belongs to

Class labels provide the correct category for each training image, such as cat or dog. In supervised learning, the model learns to associate visual features with these known categories and then classifies new images. Therefore, option A is correct. Option B concerns changing image size, whereas labels identify the image category.

Correct answer: A. To show an object's location and extent in an image

A bounding box is a rectangular region drawn around an object to indicate its location and approximate extent. Assigning one class to the whole image is image classification, not object detection. Exam tip: object detection generally provides both the object class and its bounding box.