Transferable Architecture For Scalable Visual Quality Inspection Stopper

Developing a computer vision and deep learning-based model for stoppers of visual quality inspection devices in the manufacturing industry that can work in a dynamic environment with higher accuracy and better generalization capability. Reality Current computer vision models for

Project Title

Transferable Architecture For Scalable Visual Quality Inspection Stopper

Project Area of Specialization

Artificial Intelligence

Project Summary

Problem Statement

Ideal

Developing a computer vision and deep learning-based model for stoppers of visual quality inspection devices in the manufacturing industry that can work in a dynamic environment with higher accuracy and better generalization capability.

Reality

Current computer vision models for stoppers of quality inspection devices work in specific environments in which they are trained but the performance of these models degrades and don’t generalize well in case of physical changes in environmental conditions (light, orientation, vibrations, etc). This degradation yields to lower service level agreements delivered by the quality inspection companies to manufacturers and OEMs.

Consequences

Service level agreement provided by the quality inspection companies to the manufacturers degrades because of the lower generalizability of deep learning models for the stopper stage of quality inspection. This directly impacts the level of automation and quality inspection performance of inspection systems.

Proposal

This project will improve model accuracy and generalizability of stopper stage of quality inspection under dynamic environment constraints using learning transfer-
able architectures.

Executive Summary

This project will study and develop a deep learning model using learning transferable architectures that will perform object detection for the stopper stage of the quality inspection systems. This approach will yield higher accuracy and better
generalizability in dynamic environments constraints

Project Objectives

Industry Objectives

This project will be done with the collaboration of the University of Engineering and Technology Lahore and Germany based Startup named “FototNow”. FotoNow is
a mobile-based Autonomous Visual Inspection and Fault Detection System that is easily deployed and scaled in manufacturing. By placing the capabilities of AI and Deep Learning at the fingertips of virtually any user, our solution enables Quality Inspection above human-level accuracy at
any point of the manufacturing stage with 100 The product FotoNow is using has certain shortcomings and inefficient in the visual inspection when the environment changes. The accuracy of FotoNow is decreased when it works in dynamic environments. We will improve the shortcomings by optimizing their product so that it can work in dynamic environments. It will benefit the inspection industries in inspecting across different environments without worrying about the environmental changes of the learning model.

Research Objectives

The research objective is to research and develop a neural network for the stopper of visual quality inspection by learning transferable architectures and propose an optimal algorithm for the visual inspection to work in dynamic environments with
higher accuracy and better generalizability.

Academic Objectives

• Literature review for a stopper of quality inspection systems.
• Neural Architecture Search for Stopper
• Optimizing model in term of time and performance
• End to End Application Development
• Final Presentation and benchmarking

Project Implementation Method

Development Methodology

The development methodology will be a combination of Agile Methodology for the development team in which after every sprint the development team will provide a working product of the application and the research team will work on the
shortcomings in the current solution and finding an optimal solution. After the completion of work by both of these teams, the product by both of these teams will be merged together. This will be the final output of the project and will be available for testing.

Key Milestones and Deliverable

Following will be project outcomes for this final year project,
• InDepth Literature review for neural architecture search algorithm for image classification module of hardware stopper for quality inspection
• Development and implementation of neural architecture search (NASNet, etc) based on provided dataset and benchmarking with state of art algorithms

• Optimization and porting NASNet algorithm from the previous step into mobile frameworks (CoreML, TFLite, etc)

• Deploying an optimized model into a smartphone app along with microservices-based cloud architecture.

• Documentation and possibility of publishing research paper and final presentation.

Project Activities

Following are the main activities that will be performed for the project:
• Literature Study related to the project
• Research work for finding an optimal solution
• Application development
• Algorithm optimization after the completion of Research
• Merging the application and the algorithm
• Final Product Testing

Benefits of the Project

Direct Customers/Beneficiaries of the Project

There are many industries where the FotoNow application can be used for quality assurance with just a mobile phone camera. The major are mentioned below:
• Automotive manufacturers
• Electrical Appliances manufacturers
• Heavy-Duty Vehicles manufacturers
• Aircraft inspection industry
• Construction
• Healthcare
• Packaging and Food industry

 

Technical Details of Final Deliverable

Outputs Expected from the Project

The output of this project will be a mobile application to develop a scalable visual inspection stopper system on the basis of learning transferable architectures

Final Deliverable of the Project

Software System

Core Industry

Manufacturing

Other Industries

IT

Core Technology

Artificial Intelligence(AI)

Other Technologies

Cloud Infrastructure

Sustainable Development Goals

Decent Work and Economic Growth, Industry, Innovation and Infrastructure

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com