Activity areas

Artificial Intelligence

The area of Artificial Intelligence is certainly wide, articulated and complex. Always guided by our Vision, our Mission and in particular by the concept of excellence and expertise, ELFO has first focused its R&D activities and then has implemented solutions in some areas of AI. On this page, we report those in which a particularly significant competence has been developed and in which ELFO has already had the opportunity to express itself successfully.


Object Detection

Object Detection allows the identification and the positioning of one or more objects in an image. Object Detection is used in contexts where the position of the object is a significant part of the problem and the image may contain many disturbing elements of no interest. There are many areas of application, for instance the identification of an object running on a conveyor belt.


Image Classification

Image Classification allows the images ranking by assigning one or more tags. Unlike object detection, it is static and does not allow to identify the position of two different objects within the image. Image Classification is used when it is necessary to entirely distinguish an image from other ones, assuming that there are no overlaps between images. As an example, it can be used when it is necessary to define whether a picture is a representation of one product A or another product B, assuming that there are no pictures with both products together.


Anomaly Detection

Anomaly Detection allows the identification of anomalies within large datasets, without the need for a manual definition of threshold values, but leaving to the algorithm the adaptation and the learning of what is a normal condition and what is unusual. It is often used for the identification of anomalies within large data streams (events/alarms/...), coming from IoT systems, management and monitoring systems and, in general, from any source able to communicate over a network (in Industry 4.0 logic). The relevant advantage of this technology is that it is an unsupervised ML.


Form recognizer

Form recognizer allows automatic information analysis and extraction from structured documents. The model can learn any documentation structure starting from few examples. Unlike previous technologies, the system "understands" the structure of the information and detects the information itself within the documents in a non-positional way. This means it can extract information from models derived from similar documents even when the information is structured in a significantly different way.


Classification - Suggestion

Classification techniques designed to propose suggestions can be used in all predictions scenarios based on previous data analysis, to help the user in the activities execution (e.g. forms compilation). The system can autonomously learn from the user’s behavior, adapting to the conditions that change while the system is used during its life time.



Based on the use of historical data, this allows a specific value trend identification, thus providing future estimation. Since it’s connected to measures, it can be applied to everything that can be measured. Main uses are failure predictions as consequence of signals/alarms recording, forecast of spare parts orders for worn components, raw material requirements, sales forecasting, etc.


Collaboration con Teams Chat

Collaboration is faster and more effective. Using auto-created Microsoft Teams chats, approval requests can be submitted, approved or rejected, with a single click.


Approval flow with Bots

Possible integration with external IDPs, resending the instructions in case the user is not logged in. Information are presented according to the status of the approval flow.