Computer Vision & ML
Computer Vision & Machine Learning
What is Computer Vision and how can it improve our lives?
Eagles are universally known to have the best long distance vision, while hawk’s color vision is second to none, and Owls have the most powerful night vision. These birds have extraordinary hardware built into their vision.
Computer vision allows machines to detect an image, analyze the object and identify what that image contains with accuracy. Artificial Intelligence is a robust field that helps optimize performance in numerous industries. The key to a powerful AI system is having a well-trained Machine Learning model.
A small list of industries utilizing Computer Vision:
● Automotive
● Healthcare
● Robotics
● Manufacturing
● Face Recognition
- Automotive
Computer vision is a factor for autonomous vehicles today. The technology can be used to recognize objects on the road, create 3-D maps, detect lane lines and drive in low light. In 2021 Tesla announced that it will rely on computer vision rather than lidar and radar for its new cars. The company’s chief AI scientist stated that the deep learning system is “a hundred times better than the radar.”
- Healthcare
The healthcare industry has implemented computer vision to improve patient outcomes and drive operational efficiencies. Currently the best known computer vision application in healthcare is to analyze images of scans, both to detect abnormalities in an individual and to identify patterns across thousands of scans that may inform doctors about a certain condition. Computer vision seeks out patterns that the human eye can’t pick up.
AI systems demonstrate more accuracy than human radiologists when searching for signs of cancer in mammograms, reducing the number of both false positives and false negatives.
A camera powered by computer vision can detect when a provider has forgotten to sterilize a tool or left a foreign object in a patient during surgery and subsequently notify them of the mistake.
- Manufacturing/Robotics
Cameras placed over the production line can detect these defects and alert the manufacturing workers in real time, which helps ensure quality standards are being met. Amazon Scout is an electric delivery system that delivers packages to customers safely using autonomous delivery devices. Scout launched testing in Snohomish County, Washington. The Scout system has expanded to Atlanta, Georgia, and Franklin, Tennessee.
- Face Recognition
Face recognition works by inputting the image into the database before being ready to use. Once the image is processed, it converts it into an analogue and gets saved in the database.
Face analysis is a technique in which facial features like the length of the nose, lips, size between the eyebrows, size of the forehead, basically all the features of the face are mapped.
These features are recorded and saved in the system. Once the database preserves the image, it is detected by the camera.
The system must have seen the image before and saved in the database otherwise it will not recognize the individual.
History
● Computer Vision
● Neural Networks
● Deep Learning
Neural networks
In 1944 by Warren McCullough and Walter Pitts, moved from University of Chicago to MIT in 1952 as founding members of what’s sometimes called the first cognitive science department. The neural nets described by McCullough and Pitts in 1944 had thresholds and weights, but they weren’t arranged into layers, and the researchers didn’t specify any training mechanism. What McCullough and Pitts displayed was that a neural net could compute any function that a digital computer could. The result was actually neuroscience not computer science. The discovery suggests that the human brain could be thought of as a computing device.
Deep Learning
1958 Frank Rosenblatt invented the first artificial neural network. Called Perceptron, it was intended to model how the human brain processes visual data and learned to recognize objects.
Computer Vision
Lawrence Roberts is generally considered the father of computer vision. In his doctorate thesis in 1963 at MIT. Pattern recognition and dimensional transformation (2-D, 3-D ,4-D) was obtained.