Synthetica transforms computer vision in manufacturing as we’ve never seen it before.

Modern manufacturing environments seek relentlessly higher levels of automation, resilience, and energy efficiency. Artificial Intelligence applications have become an inseparable part of the manufacturing sector retrofitting previous automation, anticipation practices and turning model-based manufacturing to data-driven. Computer vision, the cognitive part of AI that involves image processing and decision-making, has provided radical advancements in automation, assistance and guidance to robotic operations, quality inspection and operator training. SYNTHETICA, respecting the contribution of computer vision in manufacturing, has developed an end-to-end vision solution with applications in quality inspection, product identification and documentation and training. To achieve so, SYNTHETICA brings together AI-driven computer vision and synthetic data; a novel approach that replaces traditional, time-consuming data collection with simulation-based data generation.  

 The keys: Synthetic data and computer vision  

In order to apprehend the “vision” of SYNTHETICA, one has to first acknowledge the individual ingredients that comprise it. Computer vision is a field of artificial intelligence (AI) that enables computers and systems to derive meaningful information from digital images, videos and other visual inputs and afterwards, take actions or make recommendations based on that information. If AI enables computers to think, computer vision enables them to see, observe and understand. When a machine is trained to export a certain output (Y) given a certain input (X), then that specific application falls to the Machine Learning sector. 

The Machine Learning applications of Computer Vision that SYNTHETICA implements, currently face several challenges, towards industrial implementation, that are:  

a) overfitting and biased data, 

b) the need for automatic labelling of the dataset, 

c) high cost of data generation and often 

d) complete absence or unavailability of a dataset that will effectively treat the encountered challenge. 

These obstacles usually appear on the collection of real data for ML applications. In order to overcome these challenges, the SYNTHETICA solution implements synthetic data or in other words “in-silico” data that is data generated by software simulations. In the case of computer vision, the data of interest are images and therefore, a framework for generating synthetic images has been orchestrated. 

This framework renders data generation cost-effective or even minimum, eliminates privacy issues, accelerates the data collection process, and offers automatic labelling of the data that is usually extremely time-consuming. Most importantly, the SYNTHETICA data generation pipeline develops datasets and encounter challenges that the end-user hasn’t even yet created or seen before, as it requires no interaction with the real environment. 

Dataset generation is followed by Neural Networks training that in essence represents the brain of the machine and will provide the necessary decision-making and object detection actions that the end-user will take advantage of. 

In the factory, vision sensors and industrial cameras comprise the eye of the solution and are responsible for image acquisition and environmental awareness needed for the computer vision models to work on.  

Figure 1: SYNTHETICA general architecture

What does SYNTHETICA offers? 

SYNTHETICA is an end-to-end, Industry4.0 solution offering its end-users both the parts of dataset creation and model training and essentially, the final part of object detection and decision-making. Decision making and feature recognition are the aspects that provide the solution’s product value, either it is classifying manufactured products, detecting product instances or detecting special features (edges, faces, holes etc.) for post processing tasks.  

General offered features: 

  • Image dataset creation and AI model training available and adjustable by a user-friendly User Interface (suitable for both inexperienced and experienced employees). 
  • Computer Vision algorithm training, validation and testing. 
  • Data storage 
  • Legacy system integration (ERP, CMMS or SCM). 
  • Operator’s User Interface indicating the results of object detection testing and feature recognition tailored to the customer needs. 
Figure 2: SYNTHETICA web application

SYNTHETICA  supports end-to-end solution integration: 

  • Sensing system installation ensuring first time operation.  
  • First synthetic data and AI model generation and installation: SYNTHETICA generates the AI models and synthetic data required for the first installation.  
  • 3D object generation: The creation of missing 3D CAD objects for AI model and training (Optional)  
  • Sensing system maintenance and reassurance of the camera’s proper image acquisition in terms of zoom and lighting 
  • SYNTHETICA suuports the investigation of environmental changes and the need for additional lighting. 
  • Legacy system adaptation and maintenance: SYNTHETICA ensures the interaction of the platform and company legacy systems. Frequent changes in the company’s legacy systems may demand new adaptations that we also take care of.  
  • Transfer or expand the solution: SYNTHETICA takes care of expanding the solution involving potential for further generation of training data and AI models. The developer’s interface facilitates this exact scenario. 

SYNTHETICA sectors of application: 

  • Steel industry: documentation and counting of steel bars and rods. 
  • Aluminum industry: product identification of aluminum extruded profiles, quality inspection in terms of tolerance adherence 
  • Robotics assistance: product identification and real-world coordinate calculation for automating robotic tasks (e.g., pick and place, milling, drilling, joining). 
  • Food industry: quality inspection in terms of defects, improper labelling, missing or erroneous features. 
  • Packaging industry: quality inspection 
  • Additive manufacturing: quality inspection in terms of tolerance adherence and product identification for post processing tasks 


Boosting RIS countries and manufacturers 

The SYNTHETICA solution is being validated by two industrial end-users in two pilot cases. The first pilot involves its deployment in ALUMINCO S.A., an aluminum industry with main purpose the type and number recognition of aluminum profiles. Founded in 1982, ALUMINCO started as a pioneer in the products of cast aluminum of classic designs for every architectural application. Nowadays, it is one of the leading Greek Aluminum companies and exports to more than 60 countries worldwide. SYNTHETICA, as a Greek-established solution, pays attention to boosting RIS countries towards the adoption of I4.0 technologies in manufacturing and has apprehended the manufacturing problems of ALUMINCO. 

Current practices in the documentation task of aluminum profiles involves the manual counting the produced parts by operators that in many cases have to count as many as 400 profiles of very small size in one production batch. “With SYNTHETICA we will automate this verification step, saving up to 2 productive human hours in each workstation.” says George Nousias, production manager of ALUMINCO. 

Figure 3: Aluminum Profile Detection and counting using SYNTHETICA solution

However, SYNTHETICA doesn’t stop there. Italy as a RIS country with a wide manufacturing ecosystem was considered a very attractive destination for showcasing the solution’s benefits. In collaboration with Prima Additive s.r.l., one of the few manufacturers and distributors in the world to offer the best laser machines for additive manufacturing (namely Powder Bed Fusion (PBF) and Direct Energy Deposition (DED)), SYNTHETICA implements its computer vision solution towards the automation of product recognition in post processing tasks. With their customers covering a wide variety, ranging from individuals to 3D printing farms executives or industrial end-users who integrate 3D printing to their production line, the link before and after the additive manufacturing process is very important. Consequently, the additive process can be followed by other post-processing tasks or the need to identify important features of the part for documentation purposes, especially in tasks of high batch size and product number. In this occasion, where part recognition and documentation are manually performed by operators, SYNTHETICA appears a great benefit towards automating the production curriculum. SYNTHETICA is therefore, exploited as an accessory of the commercialized additive machines of Prima Additive requiring absolutely no further equipment installation and working with the machine’s preinstalled cameras. 

“With SYNTHETICA, our customers can automate the product identification and feature recognition of additive manufactured parts, increasing production rates by 15%” says John Stavridis, additive applications manager of Prima Additive. 

Figure 4: Feature recognition of dental, additive manufactured parts using SYNTHETICA Solution

What’s next? 

SYNTHETICA envisions making quality inspection and object recognition a reality in manufacturing environments. With a wide variety of applications and proven consortium expertise, SYNTHETICA targets to be successfully commercialized, starting from the Greek and related RIS sector, and expanding to the rest of the Europe and the globe. And the team has achieved so with the aid of the activity’s Business Owner, CASP S.A., a software and consulting company, based in Athens, Greece, working closely with European and local industry as well as with Universities and Technological institutes, on developing and marketing Business Solutions for a wide range of industrial problems such as ERP and SCM systems, Production Scheduling, Quality Engineering and AR/VR vision applications. 


The activity is orchestrated and coordinated by the Laboratory for Manufacturing Systems and Automation, belonging in the University of Patras, Greece, under the supervision of Professor George Chryssolouris and is oriented on research and development in cutting edge scientific and technological fields such as Manufacturing Processes Modelling and Energy Efficiency, Robots, Automation and Virtual Reality in Manufacturing and Manufacturing Systems.