Office

Via Cairoli 1/4
16124 Genoa – Italy
ph. +39 010 8970500

Computer Vision: technology to help enterprises

Computer Vision is one of the most promising solutions in the field of Artificial Intelligence and more and more companies are relying on this technology. But what is it all about?

What is Computer Vision

Computer Vision refers to a field of Artificial Intelligence that enables computers to derive meaningful information from visual input. The main purpose of this technology is to reproduce human vision, meaning not only as the ability to recognize subjects within a single or sequential image, but also to extract useful information for processing. It is, in essence, the ability to reconstruct a context around an image, giving it meaning.

 

 

 

How Computer Vision works

To learn to ‘see’, Computer Vision systems need to be trained with a large amount of images, which will build the dataset that will make the system intelligent. Computer Vision analyses data several times until it is able to recognize the images. To arrive at this final result, a special type of Machine Learning, the convolutional neural network (CNN), which extrapolates the data to obtain its own reading, is used.

 

Thanks to advances in Artificial Intelligence, Computer Vision has refined its accuracy. This is possible because the data we use and generate are constantly increasing: the more data we use, the better the machine’s performance will be.

 

The Computer Vision tasks

There are four different Computer Vision tasks:

  • Image classification, by which the subject’s class is recognized.
  • Object detection, which is the identification of objects within the image and their spatial location.
  • Semantic segmentation, which identifies similar objects in the image that belong to the same class at pixel level.
  • Instance segmentation, which identifies the different instances in the image and their boundaries at a deeper pixel level.

 

In addition, two more entries have been added to the task list in recent years:

  •  Data generation: by learning the distribution of a dataset using approaches such as GAN (generative adversarial networks), it is possible to generate new images that look real and can be used in new datasets. This overcomes the limitation imposed by Deep Learning on the amount of data needed to train a model.
  • Domain adaptation: GAN can be used to transform images from a source domain to a target domain. This is very useful for capturing the performance of networks on different tasks without having to annotate new data. This includes applications such as Deepfakes, a technology that allows the image of one person to be replaced by another within a video.

 

Applications of Computer Vision

Computer vision systems have numerous applications, both in the field of smart surveillance and in the diagnostic analysis of telemedicine. Safety at work also makes use of such software, both for predictive maintenance and safety, and for monitoring products to check their quality and detect defects.

 

If Artificial Intelligence allows a computer to think, Computer Vision allows it to see, extending its range and applications.