Artificial Intelligence has now fully entered into our daily lives: whether it’s to protect us from spam or to help us choose which movie to watch or song to listen to, advanced algorithms play an increasingly central role in our activities. But what is artificial intelligence really capable of doing? What are its limits? Here is an overview of Computer Vision, or Artificial Vision. This particularly interesting and tricky discipline is usually considered a branch of artificial intelligence.
Computer vision embraces all those techniques, strategies and algorithms that are designed to provide a sort of visual intelligence to the machine, as if it were equipped with the same visual apparatus as humans.
The implications of such functionality are fascinating and sometimes futuristic, from diagnostic support in the medical field to large-scale facial recognition and autonomous driving of vehicles.
Reaching such an objective is naturally very complex; however, modern techniques of Machine Learning and Deep Learning have allowed us to achieve significant achievements in the following tasks:
The combination of these two features allows the machine to complete more complex tasks, such as object detection (recognition of the presence and the of an object at “first sight”) and object tracking (i.e. the process of following a certain moving entity in a video stream).
Other fundamental techniques for Computer Vision concern the segmentation process, that is, the subdivision of the image into the elementary objects it is composed of, accurately identifying their shapes and boundaries. The more accurate the segmentation, the more accurate the definition of the image as a whole.
These techniques are often supported by robust Image Processing algorithms, i.e. the discipline aimed at manipulating an image at the pixel level for varied reasons: to change brightness and contrast, the introduction of new objects into the image, complex editing of the image itself, and also the extraction of features of interest.
Computer Vision experts deal with a wide and complex range of topics and methodologies, but we can grasp the innovative scope and the advantages of benefitting from these technologies.
What about recruitment?
Computer Vision can also be a winner for HR: the techniques can be used to help the recruiter analyse and evaluate the candidate, speeding up the preliminary phase of reading all of the CVs.
In particular, we refer to Optical Character Recognition (better known as OCR), a software that converts a set of words identified in an image into digital text so that the text can be recognised and manipulated by the machine. If we wanted to further simplify the operation, once the letters within the image have been identified (object localisation), it is possible to classify them into the letters of the alphabet they represent (object classification) to compose a readable and editable text in any word editor.
This technology is also used within Inda, our artificial intelligence solution for the HR world: OCR is, in fact, essential to guarantee the automatic extraction of structured candidate information from their CV. Often, it is impossible to extract text using a conventional tool (the standard case is a CV in image format).
Introducing Computer Vision systems within recruitment activities has the advantage of accelerating the collection of information from CVs, whether printed or in digital format, at every occasion (events, fairs, online or on-site selections) and through a simple photograph that allows the document to be read and transformed into text.
With an adequate system configuration, taking advantage of different Deep Learning techniques to recognize any graphic elements in the CV (object detection) and interpret them through Image Processing operations is possible. For example, a photo inside the CV can be automatically extracted.
But these are not the only advantages! The application of Computer Vision systems to search and selection activities is not limited to the analysis of documents.
Face and facial expression recognition systems have been tested to analyze the emotional states of candidates during video interviews and even to monitor the attention of candidates or users, with the common aim of constantly improving a people-centred activity.
Article written in collaboration with Angelo Schiavone, Inda’s data scientist.