Computer vision & AI
Computer vision & AI
Algorithms and models to enable machines to understand visual data
We develop advanced algorithms and models to enable machines to understand and interpret visual data. From image recognition and object detection to scene understanding, our expertise in image processing, pattern recognition, machine learning, and deep learning allows us to deliver intelligent systems that make informed decisions in the visual environment.
Computer Vision
Enable computers or machines to understand visual data
Computer vision refers to the interdisciplinary field of study that focuses on enabling computers or machines to gain an understanding of visual data, similar to how humans perceive and interpret visual information.
It involves developing algorithms, techniques, and models that allow computers to extract meaningful information from images or videos, analyze and interpret the content, and make decisions or take actions based on the acquired knowledge.
Computer vision encompasses a wide range of tasks and applications, including image processing and restoration, image recognition, object detection and tracking, scene understanding, image segmentation, 3D reconstruction, motion analysis, and more. It involves processing and analyzing visual data using various techniques such as image processing, pattern recognition, machine learning, and deep learning.
By harnessing computer vision, machines can effectively perceive and comprehend the visual world, enabling them to perform tasks such as autonomous navigation, visual inspection, surveillance and monitoring, safety, robotics, automation/quality inspection, entertainment, retail, fashion, nautical, biomedical, and many others.
In summary, computer vision aims to equip machines with the ability to understand and interpret visual data, mimicking human visual perception and enabling them to interact and make informed decisions in the visual environment.
Artificial intelligence
AI aims to equip artificial systems with reasoning and decision making
Artificial Intelligence (AI) became in recent years the “hottest” technology trend, which is still running actively. Differently from its popular literal meaning – AI is trying to replicate human intelligence – AI aims at designing computational systems with intelligent capabilities, much more than less, associated to the processing of perceptual data such as images/video, sound/speech, or text/language. In general, AI aims at equipping artificial systems with reasoning and decision-making abilities to support, or also replace in some cases, humans in performing a given task.
Nowadays, AI is most of the time a synonym of Machine Learning (ML) or Deep Learning (DL), the latter characterized by the adoption of a specific class of methods (i.e. deep neural networks). It is exactly Deep Learning that made AI so popular, since this type of technique proved to be extremely well performing in a variety of applicative tasks, especially those associated to vision and speech/language.
AI/ML/DL can efficiently cope with tasks such as visual object detection and classification, scene segmentation and recognition, automatic speech translation, retrieval, image/video captioning, multimedia content generation, image and video textual description, prediction (of actions, of stock exchange, etc.), recommendations, and many others. In doing so, it does not relate to complex rules to be manually designed, but it finds and embeds such rules automatically by training, that is, learning directly from data.
Embedding
Embedding is the process of adapting CV and AI algorithms to embedded systems
In this context, the term “Embedding” refers to the process of adapting Computer Vision and AI algorithms to embedded systems, which are hardware-optimized to achieve the necessary performance levels while minimizing power consumption, size, and cost.
Many Computer Vision and AI systems are initially developed on conventional computing platforms like PCs or servers, often equipped with GPUs, and then need to migrate to embedded architectures to achieve otherwise prohibitive performance/cost targets. Conversely, embedded architectures impose more constraints on designers compared to conventional solutions, demanding in-depth knowledge of both the problem being solved and the architecture itself. Furthermore, Computer Vision and AI algorithms present considerable design challenges due to the large volume of data to be processed and the often real-time performance requirements.
Achieving an efficient embedding of Computer Vision and AI algorithms involves two main aspects: firstly, designing custom accelerators that are optimized for executing the algorithms, and secondly, restructuring and approximating the algorithms themselves to suit the computing architecture while meeting the required Key Performance Indicators.
Services
Computer Vision and Artificial Intelligence services offered by EVS
EVS specializes in providing cutting-edge solutions in the fields of Computer Vision and Machine Learning/Deep Learning, with a specific focus on Computer Vision. Our extensive expertise encompasses a wide range of tasks, from fundamental image filtering, restoration, and segmentation to advanced operations like scene interpretation, classification, and recognition.
We possess comprehensive competencies in understanding visual data, extending beyond traditional image analysis to encompass multimedia analysis, where we seamlessly integrate sound data analysis and other data modalities, such as text or language.
Furthermore, decades of cumulative expertise in embedded systems and programmable logic development allows our research and engineering teams to collaborate closely for building highly optimized embedded software and hardware systems, accurately crafting the algorithms stack, neural network models and accelerators in lock-step.
Feasibility analysis
EVS provides in-depth feasibility studies, thoroughly assessing the technical viability of potential Computer Vision initiatives. We critically evaluate different technological approaches to ensure a solid foundation for a project.
Algorithms design
We leverage our research expertise to select state-of-the-art (SOTA) algorithms or design ad-hoc algorithms from scratch. Our engineering expertise ensures each algorithm is optimized to deliver optimal results and validated on real world data.
Dataset construction and annotation
We meticulously create and annotate datasets for AI model training and can support our customers in defining rigorous data acquisition protocols to ensure the quality and integrity of the data, setting the stage for successful machine learning outcomes.
Model design
At EVS we craft AI models with precision, employing a variety of techniques from Machine Learning to Deep Learning. We apply robust training methods such as domain adaptation and generalization to deploy reliable AI systems, even with limited data availability.
Object Detection and Tracking
Relying on our computer vision proficiency, we offer object detection and tracking services. Our software solutions accurately identify and track objects in real-time from image or video data, providing valuable insights and applications in retail, surveillance, and autonomous systems industries.
Action Recognition and Activity Monitoring
Our computer vision techniques allow us to offer action recognition and activity monitoring services. We develop software solutions that detect and analyze human actions or activities from video data, providing valuable insights in sectors such as sports analytics, healthcare, and surveillance.
3D Reconstruction and Modeling
Leveraging our computer vision skills, we offer 3D reconstruction and modeling services. Our software solutions reconstruct three-dimensional models from images or videos, finding applications in industries such as automotive, entertainment, etc.
Model optimization
Our team excels in refining AI models for enhanced performance and efficiency. We employ techniques like model compression, quantization, and optimization to deploy AI solutions on resource-constrained devices, ensuring low latency and high efficiency.
Proofs of concept and rapid prototyping
EVS brings ideas to life through comprehensive proof of concept development and rapid prototyping. We validate the practicality of Computer Vision and AI applications in real-world scenarios, providing valuable insights for further refinement.
Software development
We engineer robust software solutions powered by Computer Vision and AI technologies. With the use of a wide range of different technologies our team ensures these solutions are tailor-made to meet your specific needs.
Embedding
We specialize particularly in the design of FPGA/ASIC solutions to accelerate Deep Learning inference and Computer Vision algorithms. Our embedded solutions are optimized to deliver performance, efficiency, and reliability.
Technologies
Technologies we use
This brief compilation showcases our expertise, highlighting a diverse array of technologies that we employ in our daily operations to craft effective solutions and attain desired outcomes.
Deep learning frameworks
Pytorch
TensorFlow
TensorFlow Lite
ONNX
Keras
TorchScript
Deep learning frameworks
Pytorch
TensorFlow
TensorFlow Lite
ONNX
Keras
TorchScript
Our preferred image and media annotation tool
V7
HPC workload manager
Slurm Workload Manager
Preferred frameworks for 3D modeling
Blender
Viewer for neural network, deep learning and machine learning models
Netron
FPGA Design & Verification
AMD Vivado
AMD Vitis
Cocotb
Mentor ModelSim
VHDL
FPGA Design & Verification
AMD Vivado
AMD Vitis
Cocotb
Mentor ModelSim
VHDL
Programming languages
C++
Python
Tcl/Tk
MathLab
Programming languages
C++
Python
Tcl/Tk
MathLab
Libraries we often rely on for our solutions
OpenCV
QT
Gstreamer
ZeroMQ
OpenGL
Libraries we often rely on for our solutions
OpenCV
QT
Gstreamer
ZeroMQ
OpenGL
Operating Systems
Linux
Linux Embedded
Operating Systems
Linux
Linux Embedded
Front End
TypeScript
React JS
Bootstrap
Material UI
Front End
TypeScript
React JS
Bootstrap
Material UI
Back End
Node JS
Express JS
Flask
OpenAPI
Back End
Node JS
Express JS
Flask
OpenAPI
Database
MongoDB
MySQL
Elasticsearch
PostgreSQL
Database
MongoDB
MySQL
Elasticsearch
PostgreSQL
Cloud Infrastructure
Firebase
Microsoft Azure
Cloud Infrastructure
Firebase
Microsoft Azure
Mobile
Flutter
React Native
Mobile
Flutter
React Native
Testing
Playwright
Jest
Testing
Playwright
Jest
DevOps and CI/CD
Docker
Kubernetes
Jenkins
JFrog Artifactory
DevOps and CI/CD
Docker
Kubernetes
Jenkins
JFrog Artifactory
Source Code Management
GIT
Bitbucket
GITHUB
Source Code Management
GIT
Bitbucket
GITHUB
Case studies
Related projects
Automotive
Embedding DMS/OMS
EVS’ prowess in high performance embedding and image processing, specializing in FPGA and ASIC design technology, finds good examples in the deep strategic relationship with Seeing Machines, which extends back to 2015.
Automotive
Detecto IP core
Detecto is an IP core designed by EVS that speeds up object detection tasks in programmable logic. Detecto is optimized for AMD SoC.
Automotive
Automotive Driving Assistance Development Kit
The FPGA design of driving assistance features, such as Lane Departure Warning, Pedestrian Detection and Vehicle Detection and their integration into an automotive driving assistance development platform based on AMD Zynq-7000 SoC and Zynq UltraScale+ MPSoC devices.
Transports
Bird monitoring system
Between 2018 and 2021, EVS worked on an innovative project aimed at creating a distributed AI-based vision system for monitoring birds in airport scenarios. The team consisted of EVS, the customer, the patent holder, a research institute and a system integrator.
Transports
Vision system for sailboat docking assistance
EVS worked, in collaboration with ZF Padova, on the design of a docking assistance system for sailboats based on artificial vision. The system provides the operator with a bird-eye view of the space surrounding the boat and can perceive the obstacles on the trajectory.
Transports
DynaPCN: The People Counter
DynaPCN is a compact and autonomous passenger counting device, designed by Eurotech and based on stereo vision. It is designed for mounting above bus and train doorways and can also be used to count people as they enter or leave buildings or any area with restricted access.
Transports
Visio-Dock: docking assistance visual system
Between 2010 and 2012, EVS, in collaboration with an important Italian yacht builder, designed and developed an innovative visual mooring assistance system called Visio-Dock that aimed to improve visibility and distance perception during the boat docking operation, thus making the maneuver smoother and safer.
Healthcare & Life Science
BeQuanti
In 2013, EVS developed BeQuanti – a software specialized for the automatic analysis of leukocyte trafficking underflow – in collaboration with the LCTST lab at Verona University.
Smart Manufacturing
AST – Active Shape Tracking
EVS engineered and deployed a vision system for an OEM manufacturer of textile machines that improves the process of cutting woven labels, thereby increasing product quality and reducing waste.