The quality of an installation largely depends on the way the fiber is installed in the customer’s home and connected to the operator’s network at the Optical Network Termination Point (OPT).
An app for technicians who install your fiber
we all know that fiber is a very delicate element with a core of small dimensions that makes the two fibers have to have a perfect connection and alignment so that the signal is transmitted properly without reflections.
For its part, Peter Sandovalmanager of special projects in the Orange supply area, explains that “a good connection will avoid signal cuts and attenuations that prevent customers from being able to enjoy the services at maximum speed. To guarantee this quality, Orange has integrated an image recognition module provided by SEITECH that validates that the fusion of the fibers has been carried out correctly based on the information provided by the fusion machines used by the technicians”.
The orange operator makes available to collaborating companies an application with a module of appointment management with real-time information that makes it easy for companies to assign jobs to their technicians in a very efficient way.
The app tells the technician when to take a photo of the splicing machine by passing it to an image recognition module that is capable of recognizing the equipment model each technician uses and interprets “the indicator lights, values of attenuation, cutting angles and even detect the presence of bubbles in the fiber”.
“When it determines that the information is correct, it tells the technician that they can continue with the next step in the installation and if not, it asks them to repeat the fusion. With which we no longer depend only on the good eye and judgment of the technician, which in some cases could not be successful”, adds Pedro Sandoval. This video shows how this system works:
Deep Learning in Orange fiber
Deep learning is a machine learning subset where artificial neural networks, algorithms inspired by the human brain, learn from large amounts of data. Once the photograph taken by the fiber installation technician is received, it is passed through a neural network that has been trained to detect the model of the splicer that appears in the photograph.
Once the AI detects the make and model of the splicer, the technician will be able to start carrying out the necessary operations “for transforming and standardizing the image to correctly orient the image and correct defects in it such as brightness, shadows, etc. , after this the screen is detected in the image and it is cropped to prepare the reading”
“That already standardized image, Pedro Sandoval points out, reaches another neural network specifically trained to be capable of analyzing the errors generated by the splicers themselves. It is trained to detect splicer screens so if you pass an image of something other than a splicer like a photograph taken to a mobile phone or another image, the tool detects that it is not a splicer, so it does not allow it to continue and asks the technician to do the splicing and repeat the check”. In this way, Orange ensures that the fiber connection made in the homes of its customers is the best possible connection to navigate at maximum speed.