History And Development Of Computer Vision

From the fun and intriguing facial recognition filters on the popular apps like Snapchat, Facebook, and Instagram to the impactful programs to help vision impaired people, diagnosed cancer or improve security, has been possible and improved by the help of the image processing through computer vision. Computer vision is the transformation of data from a still or video camera into either a decision or a new representation. The introduction and development of Computer Vision has improved the ability of machines to recognize objects which in turn assist human in daily tasks or industrial processes.

Computer Vision started out in the early 1970s as an attempt to mimic human visualization of the 3D world. It was used with this purpose to develop machines and robots with intelligent behaviors. During this time, there were some early efforts to extract edges from images as well as studies on three-dimension objects in correlation to computer vision were conducted. In addition to that, quantitative approaches including feature-based stereo correspondence algorithms and intensity-based optical flow algorithms were created. Over the next few decades, computer vision was rapidly developed. Mathematic algorithms and models were heavily used during this period to process images. It was in this time that many Markov Random Field models were utilized for different image processing methods. Three-dimensional data processing was still being actively explored throughout the time. A lot of new trends, methods, and algorithms were continuously being explored as well as some old ones since the beginning such as image segmentation were still being improved. One of the big break through around this time was a linear algebra-based system called Eigenfaces developed by mathematicians Michael Kirby and Lawrence Sirovich at Brown University and later on was put into further development by computer scientists Matthew Turk and Alex Pentland at the prestigious MIT. The system was the first automatic facial recognition, and even though it still had a lot of constrained, was a big breakthrough. During the 2000s, different government agencies such as The Defense Advanced Research Projects Agency (DARPA) and the National Institute of Standards and Technology (NIST) were working on different projects to support commercial uses as well as law enforcement uses of facial recognition. It was said that 9/11 was probably the biggest reason for pushing the development and innovation for facial recognition since the US failed to identify enemies. Around late 2000s, early 2010s, object recognition technologies exploded across the different social media platforms. Starting with Google ability to reverse search image, cameras being able recognize familiar faces, Facebook, Instagram, Snapchat became popular with the ability to recognize faces on tagged photos, detected your facial features and added filters to it. However, computer vision truly took a big step and attracted many developers with the introduction of convolutional neural network (CNN).

Machine learning is a field of computer science that uses data and math algorithms to train the computer without having to code specifically what the computer need to do. CNN is a branch of machine learning that uses artificial neural network and multiplayer of kennels to analyze image. The field took a sharp turn in 2012 when AlexNet, a CNN that uses GPU, won the ImageNet Large Scale Visual Recognition Competition (ILSVRC). It was able to show that with the implementation of GPU, the CNN was trained effectively for object recognition, marking the new era for computer vision. In 2015, just three years after AlexNet was introduced, ResNet won the ILSVRC with a 3.6% error rate compared to human error rate of 5-10%. Today, computer vision in general and object recognition to be more specific is used widely in various real-world application with great prospect for the future.