Computer Vision - The past and the present

Computer Vision - The past and the present

Artificial Intelligence or in short AI happens to be that one game-changer that has all the means to alter the world of technology and innovation in gigantic proportions. Here, we are looking into one of the most powerful and captivating facets of AI which is Computer Vision, something that has permeated in our lives but most of us do not have a clue about it.

Computer vision involves mimicking the intricacy of the human visual system by machine and identifying and processing images and videos in the same way that humans do.

Credits go to advancements and innovations in artificial intelligence, deep learning, and neural networks, which has propelled the field across multiple successive levels and even surpassed humans in tasks like detection or labeling objects. Data generation is another driving factor behind such a rapid growth of computer vision, as we generate large amounts of data every day.

The Past

Although computer vision has tonnes of applications today which are growing exponentially, it first appeared in the 1950s. After almost two decades of experiments, it was in the 1970s when it witnessed commercial usage which was to distinguish between typed and handwritten text.

Computer Eye

So, what prompted such an experiment?

It stems from a question related to Neuroscience which is how do our brains work.

Let us climb up from the basic purpose of computer vision. It is all about recognizing similar features. Hence for a machine to understand such pictorial data, it needs to feed on thousands to millions of images. The next step is to run them through specifically made software processes, or algorithms, that prompts the computer to clamp down and identify similar patterns in all the entities.

Computer Metrix Human Head

Quite revolutionary for its time but it was a painstaking task and required lots of coding and other efforts by the creators and their operators. Also, this was before the advent of deep learning, so the scope of computer vision was limited.

Let us take the case of facial recognition, and here is the conventional way to do that:

  • Make a centralized database and store all the captured individual images of all the elements to be processed.
  • Specify and annotate images. Every individual image requires several key attributes and related dimensions to define the uniqueness of the image. A specific format needs to be defined as well.
  • Capture and repeat. This is repeating the first two steps again and again. Capturing images and going through required measurements and annotations once again.
  • Finally, the application is now able to compare the dimensions in a new image with the ones stored in its database and tell whether it is corresponding with any of the entities it is tracking.

Remember, in the early days, there was little automation involved and hence everything needed to be executed manually.

The Present

The advent of deep learning presented a different approach to fixing computer vision issues. It removed the aspect of manually coding every single rule into the application. It introduced specific features, like small sets of applications that can detect and identify particular attributes in images, like its dimensions. The features then apply algorithms such as logistic regression, linear regression, decision trees, etc to identify patterns, classify images and detect objects in them.

Single Depth Slice

Machine learning made it quick and it took considerably less amount of time to decipher an image. Today, we have ultra-fast processors and chips with related hardware coupled with a reliable, fast internet, and of course, the cloud servers, making the entire process extremely fast. The advantage of merging cloud with computer vision is that there is always a ready-to-use repository with millions of processed images, hence this enables us to build upon previous work rather than starting from zero. As the AI industry keeps on evolving, tasks that used to take a week take 5 minutes today. Some computer vision for real-world applications takes microseconds today.

Uses of intelligent vision

Face recognition

Facial recognition is probably the best application to come out of computer vision. Facial recognition involves algorithms that identify facial features in images and runs them through the massive repository of stored images to compare. Facial recognition first found its niche in the police department and other law enforcement bodies who use this technology to identify criminals and suspects in video recordings. Now facial recognition is found in handsets and other consumer devices that authenticate owner identities. Social media uses facial recognition to identify and verify users.

Augmented reality

Augmented reality or AR is a branch of technology that allows commercial computing devices to project upon and embed virtual entities on real-life features. Computer vision embedded in AR detects real-world objects to identify them as displayed on the device screen and place a computer-generated unreal entity.

Self-driving cars

Self-driving cars are a recent phenomenon and they are making a lot of news lately. Undoubtedly, it has been made possible by the advancements in computer vision, which is an integral part of self-driving vehicles. Computer vision helps self-driving cars to be aware of their vicinity. The smart cameras fixed on these cars capture real-time footage from different angles and feed it to the plugged-in repository. The captured footages are processed to ascertain different factors on the roads such as traffic signals, other cars, and pedestrians, lying objects, roadblocks, and road features. The self-driving car simultaneously steers its way through the roads, while avoiding hindrances, and safely making its way towards its endpoint. Self-driving is still in the build-up stage and only time will decide how far it evolves.

Self Driving Cars

Industrial fleet management

This is where intelligent computer vision finds a significant placement. From identifying and tracking all the automotive vehicles in a fleet along with their operators to monitor the routes taken, pit-stops, operator behavior, prediction of circumstance, streamlining the supply chain, reducing manual labor costs; the industrial field has benefitted immensely from computer vision and it is an ever-evolving process.

Computer vision has put a benchmark on the world of technology and certainly is an astounding feature to come out of AI. As AI keeps on evolving at a rapid pace, computer vision is becoming increasingly mainstream in our daily lives. What is more amazing is that AI is yet to reach its prime and computer vision is just the tip of the iceberg for whatever that is about to come.

iNetra-Logo-sh.png

Vehicle Tracking System

iNetra by Prescient Technologies

B507, 4th Floor, Teerth Technospace, S. No. 103, Baner, Off Mumbai Bangalore Highway, Pune 411045. Maharashtra, India
Email : contact@inetra.ai
Phone : +91-20-664 779 00

Search