> (\< Artificial Intelligence < and Visual Data > Dear children, Picking up from the previous learning Po @# That computers work differently than human brain, @ About different types of intelligence and learning styles. # How learning types and intelligence types relate? = Various applications of AT presently. # Various types of artificial intelligence. We have learnt... Welearnt that data is one ofthe main domains of Al. The other two domains of Alare Computer Vision and Natural Language Processing. We know that computers understand numbers but in real life, data is in many forms such as numbers, text, symbols, images, audio and video. As far as numbers are concerned, there is no problem in feedingthem to the computeras they are but, what aboutimagesand language? How do computers understand visuals and human language? This is where the other two domains of Al come into picture. In this chapter we shall explore computer vision. Computer Vision is the Al domain that deals with analysing visual data such as images, spatial data (images taken by satellite), video frames and live feed of datalike face, video recordingetc. How do we recognise people by their faces? Our brain is made up of millions of neurons. These neurons store the features of the faces we see and later when we meet those people, we recognise them even froma distance or by their picture. Brain is very clever in recording unique features offacesand things sothatwe remember them.How does a computer do this? For a computer, everything should be in the form of a number. When we click a photograph using a digital camera, the image captured by the camera converted into its numeric formand stored inthe memory of the device. How Computer Understands Colours? When you make a painting with water colours, you mix some colours to get new colours. You mix different quantities of water to have different colour intensity. A similar concept (but no water!) works with digital images. To understand how digital images are understood by computers, first we need to understand the basics of colours. Download Just Color Picker from annystudio.com/software/colorpicker Open Just Color Picker. Select RGB. RGB stands for Red, Green and Blue. These are 3 basic digital colours. Their combinations make other colours. Youwill notice thateach ofthe 3 colours havea range of intensity from 0 to 255. Change the intensity of Blue to 0 and that of other 2 colours to 255. Which colour did you get? Yes, Yellow. 0 to 255 is the intensity of these colours. Ifyou setall three colours’ intensities to 0, you will get black. Check itout. What will you get if you setall three colours' intensities to 255 (maximum)? White? Correct. Now try out blends of various intensities of these 3 basic colours to get different colours. There are 256"256"256=16777216 colours in RGB model. This way, computers store the colour information. There are other colour models also like Cyan, Magenta, Yellow, black (CMYK) used by printers butRGBismost common ofall. Try mixing equal intensities of all 3 colours such as 100,100,100 or 65,65,65 or 150, 150,150. Observe which shades of a particular colour are produced. How Computer Understands Images? So, colour basics are clear. All the basic colours ~ Red, Green and Blue have intensity 0 to 255. Digital images are composed of tiny elements called pixels (Picture elements). Think ofan image as a grid of several thousand tiny pixels. Here is a conceptual representation of the concept. Notice that pixel 1 isblue, pixel 446 is white and pixel 198is dark.This is just a conceptual presentation. Actual pixels are microscopic in size and there are rmillions in an image. Acoloured digital image has 3 layers -red layer, green layer andbluelayer, Imagine 3 transparent sheets ~ one red, one green and one blue, keptaboveeach other. When these 3 layers are combined, their com! intensities create the colour ofthat pixel -™ + + =f «<@ These individual layers of colours in the coloured image are called channels, ned wit —f ; aa Split Image channels Ga to pinetools.com/rgb-channels- 7 image and upload any image on it. Then, select any one channel and click on DECOMPOSE button. It will show that channel, Tf you have free image manipulation program GIMP downloaded from gimp.org then you can see the channels of an image in it. Just open the image in it and then Windows > Dockable Dialogs > Channels taconite Then remove the eye icon beside the channels in the side panel and see the SE:ity A set Secret Message of a SPY Consider the 26 images given here. Each image is a grid of 5X5 squares. Each grid is representing an English alphabet which is denoted by the grey cells in the grid. For example, L=1,6,11,16,21,22,23. Assume that each cell is a pixel. So, each image is composed of 25 pixels. Tf an Artificial Intelligence algorithm is given these figures then it can detect the pixels which are grey coloured and easily decode the alphabet. Each image has 2 parts: 1. Edge: Each grey pixel is forming the outline (edge) of thealphabet.. 2, Corner: Corner has 2 edges. For example, in alphabet A, there are 2 corners-6,1,2and 4,510. By scanning the images, AT algorithm can learn the sequence of gray cells to deduct analphabet. For example, the learning would look ike this: (156.1046, 202125) (1-6, 1-16, 21-26) (1/4116 21-28) (16,116,223) (1-6, 10,11, 15,16, 20-25) This can be decoded as the word: HELLO. Can you do this? In the code above, comma is for separate pixels and dash means contiguous pixels, For example, inalphabet H, 10-16= 10, 1,12, 13, 14, 15and 16. {7 | 8 3 {7 | 8 3 19 {17|38|15So, Al-enabled applications developed for processingimages look for the pixels that are similar in colours and intensity. They also check the density of pixels in each region of the image. Then, they checkifsimilar pixelsare ina continuous series. Allthis information help Alalgorithm to learn about the image. This is just a conceptual representation of Computer Vision, actual implementation involves a lotofcomputations butatthe base of entire conceptisthe values of pixels. Youcanuse these imagesto create your own secret messages.Applications of Computer Wi ty Translati, . petty ting Things using Computer yisio® Goto ‘thing-translator appspot.com and take the picture of any object using webcam. See how the AL algorithm identifies that object. Try a pen with and without cap. Where ever processing of images is involved, there is scope for Al-enabled visual data processing for various useful purposes. Some common uses of computer vision are: ° ° ° > Image based search. Location detection and driving directions, Processing satellite images (spatial data), Navigation of vehicles, ships, airplanesete. Object detection - faces, fingerprints, food items, stationery, fabricsetc. Cancer cells detection. Security scanners detectingand identifying facesand hidden objects. Detecting forest fires and flood situationsete. Intelligent night vision for security forces. 10. Autonomous vehicles. Smart devicessuchas security locks. Computer Visions is the AI domain that deals with the visual data, Digital images are composed of tiny elements called pixels. Coloured digital images are composed of 3 channels- red, greenand blue. Computer Visions applications are image identification, object detection, navigation, imaging in medical field, autonomous vehicles and navigation et. @Exercise A Domain: Anarea related to common features or functionality. Spatial: Related to imaging. RGB: Colour model based on the basic colours red, green and blue CMYK: Colour based on the basic colours cyan, magenta, yellow and black. Pixel: Smallest unit of a digital image that holds its colour information. Select the correct answer. 1, Whichof the following domains of AT deals with visual data? a, NLP b. cv ¢. Botha)andb) d. Robotics 2, What is the range of intensity of basic coloursinanRGB model? a, 0-285 b. 0-275 c. 0-265 d. 0-255 3, Inthe colour model CMYK, what does Kstand for? a. Grey b. Blue <. Black d. Red 4, Computers understand type of datain the formof which of the following? a. Text b. Number c. RGB d. Pixels 5. InRGB colour model, 100,0, Owill give which colour? a. Yellow b. Magenta c. Cyan d. Red 6. Tf arectangle is displayed on computer screen in green colour, then what will be its RGB values? a. 255,255,255 b. 0,0,0 c. 0,255,0 d. 255,0,255 7. Coloured digital images are composed of howmany channels? a1 b. 3 c. 255 a4 @8, Howmany edges does a corner inan image have? a2 b. 0 el a3 9, AL applications look for which of the following inan image? a. Colour b. Intensity ¢. Botha)andb) d. Noneof these 10. Which of the followings not anapplication of Computer Vision? a. Image identification b. Object detection c. Identifying voice of theuser d, Autonomous vehicle ‘Mark the following statements as True or False. 1, Datais one of the main domains of AT. For computers, all data should be in text form. Inapixel, O intensity means absence of that colour (black), Cancer cell detections easier using Computer Vision techniques. RGB model is the basic colour model used by printers, open Fill in the blanks. Images, Gray, | Numbers, Satellites, Yellow Computers can easily process The are storedin digital camerain the numeric form. Pixels (0,255, 0)and (0, 0,255) will give together, colour, Spatial data is usually provided by Pixel (100, 100, 100) will give colour. ohene Answer the following questions. 1, Name the 3 domains of artificial intelligence. What does Computer Vision domain deal with? 2. What is apixel? What do youmean by RGB colour model? 3. Howis colour model RGB dif ferent from CMYK? 4, What do youmeanby the terms edge and corner with reference to an image? 5. List any 4 applications of computer vision. www.youtube.com/edusoftknowledgeverse @