Smart factory:
Object recognition in industry
By Sigma LABS
Introduction: Object recognition
• Object recognition is an area of Artificial
intelligence(AI) connected with identifying objects
from video or image
• It uses various algorithms/models to detect objects
Purpose
• To build a flexible object and corresponding
crack(damage) detecting system with high
integrability
• The common models and algorithms are
(YOLO,GOTURN, Multibox etc.)
Problem
• Most of the production industries are based on
automated process that run with the help of scripts
• Such type of systems have low flexibility, thus
making them unable to locate, move or produce more
accurately
• Moreover, in critical situations such type of
production units are unable of make necessary
decisions without an operator
Solution
• In order to build a flexible production unit, we should
give them ability to identify objects, corresponding
damages and imperfections
• Building a flexible system will allow the unit to
control the production effectively by controlling the
quality and sorting
• Such type of system, will interact with production
unit(For example: a belt conveyor)
Process
• The cameras will inspect details for imperfections
(deformations, damages) and make appropriate
decision with the help of a pretrained model
Hardware/software
• Our image detection model will use Tensorflow,
OpenCV library with the help of Python
• Moreover, threading and multiprocessing is necessary
• The object will be inspected on several cameras at
high resolution (in order not to miss imperfections)
• In order to effectively implement object recognition,
we will use faster R-CNN developed my Microsoft
Architecture
Faster R-CNN has a Region
Proposal Network (RPN) to
generate a fixed set of
regions. The RPN uses the
convolutional features from
the the image classification
network, enabling nearly
cost-free region proposals.
The RPN is implemented as
a fully convolutional
network that predicts
object bounds and
objectness scores at each
position. Further, we will be
implementing functional
API for multitask
architecture in smart
factory
Interaction
• A PC (with installed pretrained model with the help
of OpecCV, Tensorflow) which is connected with
several cameras will seek for detail imperfections
• More over it also interacts with the
PLC(Programmable logic controller),a combination
of hardware and software which controls the
movement of conveyor
• If an imperfect shaft has found, the signal will go to
our PC through PCL and conveyor will stop and sort
an imperfect detail
Programmable logic controller.Structure
Interaction model
Team
• Scientific manager: Merey M. Sarsengeldin. Phd, associate
professor, at Satbayev University and research scholar in
Quantum machine Learning at University of Central Florida,
founder of Sigma Labs LLP and Sigma LABS Florida, Orlando,
US
• Scientific manager: Abdullah S.Erdogan, PhD, Director of
Sigma Labs Florida, Orlando, US.
• Key manager : Zholdybaev Akezhan. Satbayev University,
Sigma Labs.
• Key manager: Haluk Laman, Ph.d. Project engineer at HNTB,
Ocoee, Florida. A Civil Transportation Engineer with 7 plus
years of experience in ITS, Traffic Simulation Modeling, Traffic
Operations and Safety, Planning, Data Science and Statistics.
Member of Sigma labs
• Developers and team for fulfilling the project
http://sigmalabs.info/about.html
Financial support request
• Equipment:53 040 $
• Salary(annual):39 780 $
• Overhead expenses: 9 282 $
About Sigma LABS
Sigma LABS is a scientific organization focused on
science and disruptive technology (data science,
blockchain, IOT and their applications). For more info
please see http://sigmalabs.info
Collaboration with University of Central Florida

Presentation object detection (1)

  • 1.
    Smart factory: Object recognitionin industry By Sigma LABS
  • 2.
    Introduction: Object recognition •Object recognition is an area of Artificial intelligence(AI) connected with identifying objects from video or image • It uses various algorithms/models to detect objects
  • 3.
    Purpose • To builda flexible object and corresponding crack(damage) detecting system with high integrability • The common models and algorithms are (YOLO,GOTURN, Multibox etc.)
  • 4.
    Problem • Most ofthe production industries are based on automated process that run with the help of scripts • Such type of systems have low flexibility, thus making them unable to locate, move or produce more accurately • Moreover, in critical situations such type of production units are unable of make necessary decisions without an operator
  • 5.
    Solution • In orderto build a flexible production unit, we should give them ability to identify objects, corresponding damages and imperfections • Building a flexible system will allow the unit to control the production effectively by controlling the quality and sorting • Such type of system, will interact with production unit(For example: a belt conveyor)
  • 6.
    Process • The cameraswill inspect details for imperfections (deformations, damages) and make appropriate decision with the help of a pretrained model
  • 7.
    Hardware/software • Our imagedetection model will use Tensorflow, OpenCV library with the help of Python • Moreover, threading and multiprocessing is necessary • The object will be inspected on several cameras at high resolution (in order not to miss imperfections) • In order to effectively implement object recognition, we will use faster R-CNN developed my Microsoft
  • 8.
    Architecture Faster R-CNN hasa Region Proposal Network (RPN) to generate a fixed set of regions. The RPN uses the convolutional features from the the image classification network, enabling nearly cost-free region proposals. The RPN is implemented as a fully convolutional network that predicts object bounds and objectness scores at each position. Further, we will be implementing functional API for multitask architecture in smart factory
  • 9.
    Interaction • A PC(with installed pretrained model with the help of OpecCV, Tensorflow) which is connected with several cameras will seek for detail imperfections • More over it also interacts with the PLC(Programmable logic controller),a combination of hardware and software which controls the movement of conveyor • If an imperfect shaft has found, the signal will go to our PC through PCL and conveyor will stop and sort an imperfect detail
  • 10.
  • 11.
  • 12.
    Team • Scientific manager:Merey M. Sarsengeldin. Phd, associate professor, at Satbayev University and research scholar in Quantum machine Learning at University of Central Florida, founder of Sigma Labs LLP and Sigma LABS Florida, Orlando, US • Scientific manager: Abdullah S.Erdogan, PhD, Director of Sigma Labs Florida, Orlando, US. • Key manager : Zholdybaev Akezhan. Satbayev University, Sigma Labs. • Key manager: Haluk Laman, Ph.d. Project engineer at HNTB, Ocoee, Florida. A Civil Transportation Engineer with 7 plus years of experience in ITS, Traffic Simulation Modeling, Traffic Operations and Safety, Planning, Data Science and Statistics. Member of Sigma labs • Developers and team for fulfilling the project http://sigmalabs.info/about.html
  • 13.
    Financial support request •Equipment:53 040 $ • Salary(annual):39 780 $ • Overhead expenses: 9 282 $
  • 14.
    About Sigma LABS SigmaLABS is a scientific organization focused on science and disruptive technology (data science, blockchain, IOT and their applications). For more info please see http://sigmalabs.info
  • 15.