Part B: Subject-specific skills (theory units) · Lesson 4 of 5All lessons
AI CourseCBSE Class 10 AI (417): Unit-wise Notes & MCQsPart B: Subject-specific skills (theory units)
Lesson 4

Unit 5: Computer Vision

13 min read39 practice questions10-question quiz

Computer Vision is a domain of Artificial Intelligence that helps machines work with visual data. It enables a system to process, analyse and make sense of images and videos in a way similar to humans.

Computer Vision and its Applications

Computer Vision is the process of taking information from images, text, videos, and other visual data. It is a field of Artificial Intelligence. AI enables computers to think, and Computer Vision enables AI to see, observe and make sense of visual data like images and videos.

Computer Vision deals with extracting information from input images or videos. It uses this information to understand the visual input and make predictions. Image Processing mainly works on raw input images to improve them or prepare them for other tasks. Computer Vision is a superset of Image Processing, while Image Processing is a subset of Computer Vision.

Examples of Computer Vision include Object detection and Handwriting recognition. Examples of Image Processing include Rescaling image, Correcting brightness, and Changing tones.

The concept of computer vision was first introduced in the 1970s. Today, it is used in many areas:

  • Facial Recognition: It is used in smart cities and smart homes for security. It can recognise guests, maintain visitor logs, and support attendance systems in schools through student face recognition.
  • Face Filters: Apps such as Instagram and Snapchat use computer vision for face filters. The camera helps the machine or algorithm identify facial dynamics and apply the selected filter.
  • Google's Search by Image: Search results can be found using an image. Computer Vision compares features of the input image with a database of images and gives search results.
  • Computer Vision in Retail: Retailers can track customers' movements through stores, analyse navigational routes, and detect walking patterns.
  • Inventory Management: Security camera image analysis can help estimate items available in a store. It can also analyse shelf space, identify suboptimal configurations, and suggest better item placement.
  • Self-Driving Cars: Computer Vision is a fundamental technology for autonomous vehicles. It helps in identifying objects, getting navigational routes, and environment monitoring.
  • Medical Imaging: Computer-supported medical imaging helps physicians. It creates and analyses images and assists doctors in interpretation. It is used to convert 2D scan images into interactive 3D models for understanding a patient's health condition.
  • Google Translate App: A phone camera can be pointed at words in a foreign language. Optical character recognition sees the image, and augmented reality overlays the translation.

Everyday example: A camera app identifying an object in front of it is using the same idea of making sense of visual data.

Computer Vision Tasks

Computer Vision applications are based on tasks that get useful information from an input image. This information may be used directly for prediction or may become the base for further analysis.

The tasks used in a computer vision application are:

  • Classification: Image Classification means assigning one label to an input image from a fixed set of categories. It is one of the core problems in Computer Vision and has many practical applications.
  • Classification+ Localisation: This task identifies what object is present in the image and also identifies where that object is located in the image. It is used only for single objects.
  • Object Detection: Object detection means finding instances of real-world objects such as faces, bicycles, and buildings in images or videos. Object detection algorithms usually use extracted features and learning algorithms to recognise instances of an object category. It is commonly used in image retrieval and automated vehicle parking systems.
  • Instance Segmentation: Instance Segmentation detects instances of objects, gives them a category, and then gives each pixel a label based on that. A segmentation algorithm takes an image as input and gives a collection of regions or segments as output.

Everyday example: If a system sees one bottle in a picture, classification can name it, localisation can show where it is, detection can find such objects, and segmentation can mark its pixels.

Basics of Images: Pixels, Resolution and Pixel Value

A digital image is stored in a computer as numbers. These numbers represent pixels, and pixels are arranged in a particular way to create the picture we recognise.

Pixels:

  • The word pixel means picture element.
  • Every digital photograph is made up of pixels.
  • Pixels are the smallest units of information that make up a picture.
  • They are usually round or square.
  • They are normally arranged in a 2-dimensional grid.
  • More pixels make the image closer to the original.

Resolution:

  • The number of pixels in an image is sometimes called the resolution.
  • One way to express resolution is width by height.
  • A monitor resolution of 1280x1024 means 1280 pixels from one side to the other and 1024 pixels from top to bottom.
  • Another way is to express the number of pixels as one number, such as a 5 mega pixel camera.
  • A megapixel is a million pixels.
  • 1280 x 1024 = 1,310,720 pixels, which is about 1.31 mega pixels.

Pixel value:

  • Each pixel stored in a computer has a pixel value.
  • This value tells how bright the pixel is and/or what colour it should be.
  • A common pixel format is the byte image.
  • In this format, the number is stored as an 8-bit integer.
  • The possible values range from 0 to 255.
  • Zero is usually taken as no colour or black.
  • 255 is usually taken as full colour or white.
  • Computer data is stored using ones and zeros, called the binary system.
  • One byte is equal to 8 bits, and 8 bits give 256 values, numbered from 0 to 255.

Everyday example: Zooming into a digital photo until it looks like tiny blocks shows how pixels build the picture.

Grayscale Images and RGB Images

Grayscale Images:

  • Grayscale images have shades of gray without visible colour.
  • The darkest possible shade is black.
  • Black is the total absence of colour or the zero value of a pixel.
  • The lightest possible shade is white.
  • White is the total presence of colour or the 255 value of a pixel.
  • Intermediate shades of gray are shown by equal brightness levels of the three primary colours.
  • A grayscale image has each pixel of size 1 byte.
  • It has a single plane of a 2D array of pixels.
  • The size of a grayscale image is written as Height x Width.

RGB Images:

  • Coloured images are made from three primary colours: Red, Green, and Blue.
  • Different intensities of red, green, and blue combine to make colours.
  • An RGB image is stored in three channels called the R channel, G channel, and B channel.
  • Each plane has many pixels.
  • Each pixel value in a plane varies from 0 to 255.
  • The three planes combine to form a colour image.
  • In an RGB image, each pixel has a set of three values.
  • These three values together give colour to that pixel.
  • When the R, G, and B layers are seen separately, they may look like grayscale images because each pixel value shows intensity from 0 to 255.

Everyday example: A colour on a screen is formed by combining red, green, and blue values for each pixel.

Image Features

In computer vision and image processing, a feature is a piece of information that is useful for solving a computational task in an application. Features can be specific structures in an image, such as points, edges, or objects.

Images can give many features. These can include:

  • a blob
  • an edge
  • a corner

Features help in performing different tasks and doing analysis based on the application.

Flat areas are difficult to find and track. A flat patch can look the same at many places in an image. Edges are easier than flat areas, because they show a boundary, but exact location may still be difficult because an edge can look the same along the line. Corners are the easiest to find because moving a corner patch changes how it looks. This makes the corner unique.

Good features are those that can be found at a particular location in an image. Corners are considered good features in an image. Edges are useful too, but they are spread over a line and may look the same along the edge. Therefore, corners are good features to extract from an image, followed by edges.

Everyday example: A plain wall patch is hard to locate, but the corner of a window is easier because it has a unique shape.

Convolution and Kernels

Computers store images as numbers. Pixel values range from 0 to 255, and the value of a pixel decides the colour of that pixel. If these numbers are changed, the image also changes. This changing of pixel values is the base of image editing.

Filters used in image editing change pixel values evenly throughout an image. This is done using convolution and the convolution operator.

Convolution is a simple mathematical operation used in many common image processing operators. It gives a way to multiply two arrays of numbers, usually of different sizes but of the same dimensionality, to create a third array of numbers with the same dimensionality.

In image convolution:

  • An image array is multiplied element-wise with another array called the kernel.
  • After multiplication, the values are added.
  • I represents the Image Array.
  • K represents the Kernel Array.
  • I * K represents the resulting array after applying the convolution operator.
  • The kernel is passed over the whole image to get the resulting array.

A Kernel is a matrix. It is slid across the image and multiplied with the input so that the output is enhanced in a desired way. Different kernels have different values for different effects that can be applied to an image.

In image processing, convolution is used to extract features from images. These features can be used for further processing, especially in Convolution Neural Network (CNN).

During convolution, the centre of the image region is overlapped with the centre of the kernel to get the convolution output. The output image can become smaller because overlapping happens at the edge row and column. To keep the input and output image the same size, the edge values are extended out by one in the original image. While extending the edges, the pixel values are considered zero.

Summary:

  • Convolution is a common tool used for image editing.
  • It is element-wise multiplication of an image and a kernel to get the desired output.
  • In computer vision applications, it is used in Convolutional Neural Network (CNN) to extract image features.

Everyday example: Applying a filter to the same photo changes the output because the pixel values are changed.

Convolution Neural Networks (CNN)

A Convolutional Neural Network (CNN) is a Deep Learning algorithm. It can take an input image, assign importance through learnable weights and biases to different aspects or objects in the image, and differentiate one from another.

The deployment process of a CNN is simple to understand. An input image is given to the CNN. The image is processed through the CNN. The CNN then gives a prediction based on the label given in the dataset.

A convolutional neural network consists of the following layers:

  • Convolution Layer
  • Rectified linear unit (ReLU)
  • Pooling Layer
  • Fully Connected Layer

Convolution Layer: This is the first layer of a CNN. Its objective is to extract features from the input image. The first Convolution Layer conventionally captures Low-Level features such as edges, colour, and gradient orientation. With added layers, the architecture adapts to High-Level features too. This gives the network a fuller understanding of images in the dataset.

The convolution layer uses convolution operation on images. Several kernels are used to produce several features. The output of this layer is called the feature map. A feature map is also called an activation map. Both terms can be used interchangeably.

The feature map is useful because:

  • It reduces the image size so that the image can be processed more efficiently.
  • It helps focus only on the features that are needed for further processing.

Rectified linear unit (ReLU): After the feature map is made, it is passed to the ReLU layer. This layer removes all negative numbers in the feature map and keeps positive numbers as they are. Passing the feature map through ReLU introduces non-linearity. The ReLU graph begins as a horizontal straight line and then increases linearly when it reaches a positive number.

The ReLU layer makes colour change more obvious and abrupt. This makes edges more obvious and gives better features for further CNN layers. It also enhances the activation layer.

Pooling Layer: The Pooling Layer reduces the spatial size of the Convolved Feature while keeping important features. Two types of pooling can be performed on an image:

  • Max Pooling: Max Pooling returns the maximum value from the portion of the image covered by the Kernel.
  • Average Pooling: Average Pooling is listed as another type of pooling performed on an image.

The Pooling Layer is important because:

  • It makes the image smaller and more manageable.
  • It makes the image more resistant to small transformations, distortions, and translations in the input image.

A small difference in the input image creates a very similar pooled image.

Fully Connected Layer: The final layer in the CNN is the Fully Connected Layer, also called the FC layer. Its objective is to take the results of the convolution and pooling process and use them to classify the image into a label in a simple classification example.

The output of convolution and pooling is flattened into one vector of values. Each value represents a probability that a certain feature belongs to a label. For example, in an image of a cat, features such as whiskers or fur should have high probabilities for the label cat.

Everyday example: A CNN can process a picture in stages, first finding useful features and finally giving a label.

No-Code AI Tools for Computer Vision

No-code AI tools allow models to be trained without writing code. In this unit, the named tools are Lobe.ai, Teachable Machine, and Orange Data Mining Tool.

Lobe.ai:

  • Lobe.ai is an Auto-ML tool.
  • It is a no-code AI tool.
  • It works with image classification.
  • It allows a set of images with labels.
  • It automatically finds the most optimal model to classify the images.

Teachable Machine:

  • Teachable Machine is an AI, Machine Learning, and Deep Learning tool.
  • It was developed by Google in 2017.
  • It runs on top of tensorflow.js, which was also developed by the same company.
  • It is a web-based tool.
  • It allows training of a model based on different images, audio, or poses.
  • The input can be given through webcam or pictures.

Smart Sorter activity idea:

  • The purpose of using Computer Vision is to automate and enhance sorting processes through computer vision technology.
  • The activity uses three classes: Bottles, Cans and Paper.
  • The model is trained and then the classifier is tested.

Orange Data Mining Tool:

  • Orange Data Mining Tool is used in a real-world Classification Model: Coral Bleaching.
  • Coral can be found in tropical ocean waters around the world.
  • Coral reefs are large underwater structures made of skeletons of marine invertebrates called coral.
  • Corals are an integral part of aquatic life.
  • The purpose of the activity is to develop a classification model for early identification of coral bleaching to safeguard marine ecosystems.
  • Detecting bleached corals at early stages helps prevent aquatic life from disaster.

Everyday example: Sorting pictures into labelled groups is the same basic idea used in a no-code image classification model.

About these notes

These notes and questions were written by Syllabax from the official CBSE Class X Artificial Intelligence (417) Facilitator Handbook and the 2026-27 curriculum. This is not an official CBSE publication.

✦ Computer Vision✦ Image Processing✦ Classification✦ Classification+ Localisation✦ Object Detection✦ Instance Segmentation✦ Pixel✦ Resolution✦ Pixel value✦ Grayscale image✦ RGB image✦ Feature✦ Convolution✦ Kernel✦ Convolutional Neural Network (CNN)✦ Convolution Layer✦ Feature map✦ Rectified linear unit (ReLU)✦ Pooling Layer✦ Fully Connected Layer✦ Lobe.ai✦ Teachable Machine✦ Orange Data Mining Tool