Visiondevice LogoVision Device Logo 40years
15 October 2022


Deep Learning is a specific area of research that studies the possibilities and methodologies of machine learning. Although the Italian terminology for defining this field is Deep Learning, it is rarely used, preferring the English term.

Deep Learning: what it is and how it works

This is a branch of Artificial Intelligence that is probably at the height of its development today but which saw its beginnings around the mid-1960s, a time when the first algorithms for machine learning of certain computers were developed.

Those initially interested in this type of process were mainly Soviet and Japanese computer scientists who were able to create simple machine learning architectures that initially took a time considered too long to apply and use: four days.

The following decades saw the improvement of learning structures, which became more and more complex while, at the same time, decreasing the time for machine learning.

In 1986 the term Deep Learning was born, although its widespread use began only from the year 2000 onward. Then in recent years, following a series of excellent purchases, such as that of the company Deep Mind by Google, Deep Learning has (almost) entered the mainstream.

Neural networks foundation of deep learning

At the basis of Deep Learning are neural networks, i.e. special algorithms capable of making autonomous decisions based on certain inputs given by the system. Depending on their development and level of advancement, neural networks can improve with experience, i.e. recognise variables that cause errors and eliminate them when it is necessary to repeat the same iteration. All this is made possible by the presence of a memory and a brain capable of managing the different signals to combine them in such a way as to give a result as close as possible to reality, i.e. to what is required.

To better understand the functioning of neural networks, one has to think that they operate on certain data, which can be video or audio signals, colours, words and so on. Modern strategies have focused on algorithms of two different types, those that are more general and those that are highly targeted.

The difference lies mainly in the field of application, which can range from voice recognition to autonomous driving of means of transport, from photo and video recognition to much more, such as Machine Vision. However, it should be specified that although Deep Learning is being applied in many areas today, its potential is still very broad.

Deep Learning and automatic driving

One of the fields of application of Deep Learning that most interests the general public is probably that of autonomous driving. It is still in the experimental and prototype stages, but there are many who have developed algorithms to teach cars how to drive in the absence of a human driver.

Il deep learning e la guida autonoma

The latest prototypes presented are characterised by neural networks capable of recognising their surroundings and, consequently, of being able to move alone even in traffic. Naturally, in order to ensure maximum safety, the cars will be equipped with a series of cameras whose images can be constantly processed by software to ensure control on the different sides of the road.

From this point of view, a fundamental branch of Deep Learning involves the support of Machine Vision, which allows the various sensors and cameras to reproduce human vision, i.e. to recognise the entire context in which one moves and thus be able, via the brain, to decide on possible directions to take.

Again, sensors of various types will allow the recognition of sounds to increase the parameters that the computer must take into account to ensure safe driving. 

Unlike the first algorithms that took days to process a minimal amount of information, today's neural networks in prototype cars can handle and process over twenty billion operations per second, all learned through Deep Learning.

This information includes, of course, all those related to safe driving, such as checking the roadway, the presence of obstacles or pedestrians, recognition of adverse weather conditions and much more.  

Deep Learning in medical diagnostics

Beyond the interesting applications in the world of cars, the fields in which work is being done to improve neural networks and Deep Learning are diverse, ranging from diagnostic programmes in medicine to quality control in industrial production.

Probably medical diagnostics is an area in which the interest in the development of neural networks is greatest, since algorithms are already being used in many cases to support specialists. The success of neural networks in the medical sector is precisely due to the complexity of the human body and, therefore, the different and likely solutions of the diagnostics themselves.

These, in fact, are most often based on previous experiences that the doctor has had, lived through or studied and that enable him to make a specific diagnosis. Neural networks in this field can therefore start from an already extensive background and further improve, through deep learning, their knowledge in specific areas, just as humans do but with greater speed.

Deep Learning in quality control systems

Another area in which neural networks can make a fundamental contribution is in programmes relating to quality control systems in companies and industries. The quality of production, in fact, is in most cases recognised through sight and touch, i.e. by observing products or feeling them with the hand. After defining the minimum level of quality, then, quality controllers proceed by sorting and eliminating everything that does not correspond to the company's minimum level of quality.

Of course, in large production contexts, the possibility of checking the quality of products by means of a programme capable of recognising the characteristics of the product itself would allow a major improvement in the speed of quality control. Neural networks, as well as Machine Vision, become fundamental in this context because they allow the quality of products to be checked, possible factory defects, faults, wrong colours and much more, depending on the field in which one is operating.

However, to ensure a complete job, it is necessary to combine the recognition system with a mechanical element capable of eliminating products that do not meet the defined standard.


In conclusion, it can be said that Deep Learning today engages some of the brightest minds working in the IT sector: start-ups, companies and research centres often collaborate to try to improve what, at the moment, is no longer just a theory, but still cannot be used to its full potential in different sectors.

Contact us







    *mandatory field

    I have read and understood the notice Privacy*

    I agree with the processing of my data for marketing and commercial information purposes

    NON CANCELLARE - è per far caricare il font dal sistema
    font: LATO
    NON CANCELLARE - è per far caricare il font dal sistema
    linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram