What is Computer Vision (CV)?
CV is one of the many fields of Artificial Intelligence (AI) that allows computers or systems to ‘see’, and act based on information contained in images, videos, or other visuals. Not to confuse this with classic machine vision which is the use of existing software technologies and tools to help machines communicate data and solve real-world problems in industrial environments.
Another key difference is that CV (in most use cases) relies on a trained data set, this can be the difference between good or bad, classification, object detection, semantic segmentation, anomaly detection or instance segmentation.
How long has CV existed?
CV continues to be a big topic within the machine vision industry, though the topic is not as new as you may think. Experimentation with CV began in 1959 and by the 1970s scientists and engineers had ready the first commercial use of CV with Optical Character Recognition (OCR).
Computer Vision for Industry.
Computer Vision relies on Deep learning (DL) which is another subset of AI. DL is becoming more prominent in industrial sectors such as pharmaceutical, food & beverage, and automotive. Is CV a need or a want? This is a big question that engineers need to ask themselves as many people believe that CV/DL will ‘just work’ and do what they want straight away. Unfortunately, this is not the case and some applications require a lot more work than using a traditional classic machine vision approach.
However, there are some very good use cases for using CV. The key to a successful CV application is to have an extensive dataset for the DL algorithms, the results will only be as good as what you put in after all. With a properly trained neural network, a huge amount of time and cost can be saved for various inspection tasks.
Why use Computer Vision over Machine Vision?
Whilst traditional machine vision and computer vision have clear differences, the only real change to the system is the image processing library. One uses a DL algorithm, the other does not. The same camera, lens, and light source are often used in each approach. However, both methods have distinct strengths but are not mutually exclusive. It is even possible to use both traditional machine vision and deep learning in the same system.
What to consider?
- Application Characteristics
- Training Data
- Computing Power
|Computer Vision||Traditional Methods|