Replicating human interaction and behavior is what artificial intelligence has always been about. In recent times, the peak of technology has well and truly surpassed what was initially thought possible, with countless examples of the prolific nature of AI and other technologies solving problems around the world. Think about this: Gary Kasparov stated that he would never lose a game of chess to a computer. For a long time, this seemed like a statement that would withstand all tests. Roll on 1996, however, and IBM developed Deep Blue, a computer bot/program/application that beat the master Gary Kasparov at his own game. This was perhaps the first practical taste of the power of AI, and since then, the artificial intelligence world has well and truly been on steroids. What is Computer Vision? Computer vision is a similar step in the direction of AI supremacy. The aim that powers computer vision is to replicate human perception and understanding of an image, and then draw useful information from that image. Computer vision may seem like a technology visited infrequently by the realms of science, but the truth is, even simple concepts involved in computer vision have taken ages of research and innovation to be made available for practical use. Now that all the required bits and bytes are here, it is time for us to realize the potential that computer vision holds in solving problems, without all the noise. Back where it all started A decade-ish ago, people used to tag their friends in photos on Facebook, and the platform then recommended tagging people in future uploads, based on previous tags. Many people would not have realized this, but Facebook's ‘tagging’ algorithm/feature was one of the first and most prominent uses of computer vision and image processing. Ever since, image recognition as part of computer vision has been employed to unlock smartphones, by banks to verify transactions, and by establishments to identify wrongdoers and those who break the law, for security purposes. As a matter of fact, Computer Vision is a big part of digital enterprise in today's AI-first scenario. Surveillance and security have been two of the prime applications of computer vision that have evolved from a simple ‘tag your friends in a photo’ algorithm on a social media platform. How is computer vision solving other significant world problems, without uttering a word (literally)? Let’s find out. Helping catch criminals, 24X7 Computer vision software can be added to surveillance cameras for as low as USD 50 per month to help log license plates of cars that pass through the area. Taking the example of Rotterdam in New York, having a very small force of police officers, a police department computer logs around 10000 license plates of cars that pass in and around town. Robberies, thefts and other crimes can then be solved using the license plate tracking capabilities of the cameras around town. License plates, as we know, are unique for every car. Using the information that the mighty computer vision provides us, officers can then use the department to search for the make, model, and owner of the car. Pretty nifty and useful for detective work, wouldn’t you say? Computer vision, machine learning, and the world of fashion Let’s move on to the next big problem that computer vision solves- fashion dilemmas. Ever wondered what to wear with that new pair of jeans you bought? Can’t think of which jacket to wear with a funky blue shirt? Let AI do the work for you. (No, this is not an advertisement for any product.) AI, computer vision, and machine learning, when combined, can tell you what would look best with a denim jacket. Yes, we are not kidding, yes, this has been achieved already, courtesy StyleIt. Over time, applications like StyleIt track your wardrobe choices and preferences, and over time, learn from the data to provide users with more customized preferences and options for outfits. An app that picks out your fashion for the day- AI on steroids is quite literally splendid! Transforming Education Computer vision has been put to use in security, surveillance, and even fashion, as we saw, but the time to utilize AI on steroids to bring about meaningful change to the world is finally here. Computer vision and its role in transforming education is that of easing the assessment process. Tracking of student behavior while teaching online classes, or even tracking their engagement in traditional classrooms lays the cornerstone for computer vision’s foray into education. OpenEdx, the open-source platform has also done some really great work in computer vision. Known for helping institutes and instructors to , it would be great to have them bolster the adoption of computer vision in the education industry. launch courses using OpenEdX Performance measures like attentiveness, interest level, and concentration can be gauged by simply analyzing “the look on students’ faces”, so to speak. Educators can then use this data to modify their teaching methods, and put together customized courses or modules that students find more interesting and pay more attention toward. Tweaking the assessment and curriculum strategy based on student response is a large part of computer vision’s contribution. Another use case of the aforementioned performance measures would be assigning different groups to students, based on their interest levels. Educators can place students into groups where they are true to themselves, where students feel comfortable learning and sharing their views. Growth and peer to peer interaction between students is promoted, adding to the long list of benefits and education considerations that computer vision provides. We all know the wonders of when it comes to enabling peer-to-peer transactions and collaboration such blockchain technology making a cryptocurrency exchange. The next step Robots have been around for quite some time, now. One of the burning questions, however, that remains is that when are we going to see robots being deployed as a common sight. The reason behind the delay in deployment is that robots cannot see very well. By not being able to see very well, we mean emulating the human reaction time and behavior to what our eyes see. Computer vision and the challenges that lie within the technology have been applied well to programs, but robots seem to be the next logical step. Image processing is an integral part of computer vision To talk about the next step in computer vision, we need to go back in time. Going as far back as the late 80s, Yann LeCun, programmed an approach to computer vision that mimicked the way our brain comprehends images. Most state of the art systems work on this very approach, running different software filters over pixels and collections of pixels in images. What he did differently was devise the concept of convolutional neural networks, which differ from traditional networks in the sense that these devise software filters automatically (human input is often required for traditional systems). The next step, adding unsupervised learning to this algorithm. No instruction is provided to the system, it is just shown image upon image, and expected to apply the correct filter to arrive at the output. The correctness is judged by the resemblance of the output image with the input. Convolutional neural networks have been used in Google Street View, and as recent as the 2010s, it even caught the attention of the DARPA. The future for computer vision certainly is promising, and all developments made in the field as of now bring in a feeling of promise for the foreseeable future.
