COMPUTER VISION

"Computer vision and machine learning have really started to take off, but for most people, the whole idea of what is a computer seeing when it's looking at an image is relatively obscure"

HISTORY OF COMPUTER VISION

Computer vision(cv) has been around for over 50 years, its development began in the 1950,s around the same time when artificial intelligence gained prominence. In the summer of 1966,Seymour Papert and Marvin Minsky at MIT Artificial Intelligence group started a project titled Summer Vision Project.

The main aim of this project was to build a system that could analyze a picture and identify all objects in the image, it is interesting to note that even now CV is a complicated area with thousands of researchers across the globe trying to perfect it but it was started by two undergraduates who dared to solve it way back in the 1960,s ,in the 1970,s,David Marr ,a neuroscientist at MIT, taking cue from brain studies including cerebellum, hippocampus and cortex for human perception, set up the building blocks of modern computer vision. He is known as the father of modern computer vision.

WHAT IS COMPUTER VISION AND ITS SIGNIFICANCE IN OUR LIVES

We are constantly surrounded by pictures ,be it google or your phone or social media. People are taking selfies all the time and sharing on Instagram or Snapchat. The reason for this growth in images and videos is because of smartphones which have cameras ,and so taking a video or photo and sharing it has never been easier, and mostly because of which we see the internet is flooded with photos and videos, but soon we realized that it would be very efficient to analyze data from the photos & videos by the computer itself, soon after  Seymour Papert and Marvin Minsky at the MIT Artificial Intelligence group unknowingly introduced us to the concept of Computer Vision. It is the concept in which the computer extracts practical information via multimedia such as Photos, Videos etc. It’s an excitingly abstract concept where we teach the computer how to see & extract data via multimedia, which uses many kinds of methods, practices and algorithms. After the computer extracts the data, it processes it & uses it for different kinds of functions which has been programmed in the computer by the Computer Programmer. Some examples can be in youtube, the captions of a video are seen during the video simultaneously, it’s because the website is using methods & algorithms which extract the data from the audio of the video, process the data and show the output as the captions of the video.  Computer vision helps in marketing, by processing images automatically by identifying brand logos very fast, understanding optimal color patterns for various markets and by searching for the subject of images.CV helps companies to satisfy targeted markets in a more personalized way.

Deep learning can work in hand with CV creating powerful systems such as searching images in google, tagging of friends on social media, snapchat filters, speech to text translation, intrusion, detection systems.

The world is undergoing a fundamental digital transformation that shows no signs of slowing down. According to a report, every minute

1.Users watch 4,146,600 YouTube videos

2.Instagram users post 46740 photos

3.Snapchat users share 527760 photos

All the above data sets give us a huge opportunity to apply computer vision to derive patterns and analyze this data and develop useful applications out of it.

INTERESTING READ

Amazon go is a new kind of store with no checkout required. With the just walk out shopping experience, simply use the amazon go app enter the store, take the products you want and go! No lines,no chekout,the key underlying technology is computer vision 

CHALLENGES WHILE USING COMPUTER VISION 

One would think that with so much advancement in technology ,Computer vision is superior to human vision and can surpass it on all use cases. But this is not the case. There are still lots of constraints in cv capabilities. When compared to human vision

in situations like simple face recognition under varying circumstances of lighting, expression, additional objects/clothing on face, CV output is sometimes not correct. Some of the challenges are

1.Privacy and Ethics

The insurance industry is trying to customize motor insurance premiums by analyzing user driving behavior through surveillance

systems. On the flip side, vision powered surveillance systems can help identify certain ethnic groups and minorities which are not desirable, government needs to take appropriate steps to tackle data privacy.

2.Lack of explainability

Modern artificial neural network based algorithms are still not clear in terms of how object classification is carried out. How self driving cars are designed algorithmically and their accuracy levels in object collision avoidance,etc are still not very clearly understood leading to ambiguity in laws around it.

3.Deepfakes

Using deep learning techniques, it is possible now to create fake images and videos of celebrities impersonating them and broadcast any message or communication. Its use during election time can be damaging to democracy.

"AT AN ABSTRACT LEVEL ,THE GOAL OF COMPUTER VISION PROBLEMS IS TO USE THE OBSERVED IMAGE DATA TO INFER SOMETHING ABOUT THE WORLD"

ACTIVITY 

1.Teachable snake is an interactive web game ,the idea is to use physical buttons to control the game, the user can draw a black arrow on a piece of white paper as a controller and move the snake by turning the paper in different directions in front of the webcam 

Teachable Snake by Vince MingPu Shao - Experiments with Google

2.Auto Draw

This is an AI experiment, which identifies the pattern of the drawing that you are making and suggests images similar to these drawings

AutoDraw 

HOW DEEP LEARNING IS ENHANCING CV AND WHAT IS IMAGE PROCESSING?

Image processing is defined as the process of creating a new image, from an already existing one, either simplifying or enhancing the content. A Given CV system may require image processing to be applied to its raw input eg pre processing images 

KEY ELEMENTS OF CV APPLICATIONS INCLUDE BUT ARE NOT LIMITED TO

1.Object classification-Category of object contained in the image such as human or animal

2.Object identification-Type of object such as cat or dog

3.Object Verification-Verify if the object is actually in the image

4.Object Detection-Where exactly is the object in the image

5.Object Segmentation-Pixels specific to the object in the image                                                                                             

6.Object Recognition-Overall understanding of the object and where it is in the image.                                                            

Deep learning has fundamentally transformed CV capabilities. Traditionally, in the Pre-Deep Learning era, CV was constrained with computational capabilities, very little automation was involved and cost of recognition would be high as the database became larger, not to mention constraints on capabilities of identifying images due to change in lightning,angle,attributes-structure of image etc Deep learning provides a completely different novel approach to solve computer vision problems, with deep learning we no longer need to face problems while identifying images of different attributes as per lightining,angle of the image or even if there is any foreign object in the image which our system may find difficult to identify but with deep learning it becomes convenient and  we can program 'features'-smaller applications that could detect specific patterns in images and make the image easier to identify.

MACHINE LEARNING AND DEEP LEARNING

Machine Learning is the field of developing and understanding methods which will help the computer to ‘Learn’, whereas Deep Learning is a broader type of machine learning which imitates the way we human beings learn, which makes it powerful and efficient, alongside machine learning. Both Machine Learning & Deep Learning are required for Computer Vision as these are the methods and algorithms belonging to machine & deep learning which help the computer extract data in the efficient way possible & provide information after processing the data which was extracted. 

APPLICATIONS OF COMPUTER VISION

1.FACIAL RECOGNITION APPLICATIONS

Social media tagging, phone unlocking using face id and attendance biometric systems, law enforcement agencies use face recognition to identify criminals who change their facial features to go incognito, passport agencies, immigration authorities 

   

2.HEALTHCARE

Computer Vision can analyze health records, lab reports ,X-ray images and prescription data/clinical information much faster than doctors. It can then run algorithms against millions of empirical records and do prognosis effectively an example of this application in health care is Orlando Health Winnie Palmer Hospital for women and babies, which taps computer vision through an artificial intelligence tool that measures blood loss during childbirth.

3.MIXED REALITY

You must have seen google glass,virtual reality(VR),headsets and augmented reality(AR) apps,these applications use CV to empower mixed reality experience, they are used in education, in tourism industry for creating VR tours and in oil mining industries to capture offsite areas

4.E-COMMERCE

One of the most important function in E-commerce is automatic product categorization. Whenever a new product is made available on E-Commerce store,its aspects are automatically updated with the help of Computer Vision system ,it helps the products to go up on the virtual shelves and within the consumers reach faster 

5.MEDIA

Visual effects for film,TV(reconstruction),virtual sports replay,semantics based auto edits.

"THE GOAL OF COMPUTER VISION IS TO EXTRACT USEFUL INFORMATION FROM IMAGES. THIS HAS PROVED A SUPRISINGLY CHALLENGING TASK;IT HAS OCCUPIED THOUSANDS OF INTELLIGENT AND CREATIVE MINDS OVER THE LAST FOUR DECADES,AND DESPITE THIS WE ARE STILL FAR FROM BEING ABLE TO BUILD A GENERAL PURPOSE SEEING MACHINE'."

INTERESTING THINGS TO SEE AND WATCH