Book Free Consultation
Book Free Consultation

How Data Pilot helped a global fashion brand streamline its defect detection process

Duration

7 Months

Industry

Retail & E-commerce

Services

AI/ML, MLOps, Custom Data Solutions

Tools And Technologies

Company Background

Levi’s is a dominant player in the apparel market, renowned for its iconic denim products and commitment to uncompromising quality. The brand operates multiple product lines, including jeans, jackets, and accessories, each catering to diverse market segments from casual wear to premium fashion. Seasonal collections like Spring/Summer and Fall/Winter are crucial to Levi’s business strategy, requiring strict adherence to quality standards to maintain the brand’s esteemed reputation.

Challenges

Errors in manually detecting defects

The manual defect logging process slowed down production, resulting in delays and inefficiencies throughout the supply chain.

P

Inaccurate data entry

Reliance on manual data entry increased the risk of errors, potentially allowing quality issues to go unnoticed.

P

Supply chain delays

The inefficiencies caused by these manual processes led to missed deadlines and disruptions, impacting Levi’s ability to meet customer expectations.

P
jQuery('.plus-icon-click').on('click', function() { var contentElement = jQuery(this).closest('.et_pb_column') .find('.card-clickable .df_tc_content_container .content'); contentElement.toggleClass('show'); jQuery(this).toggleClass('active'); // Toggle the icon content if (jQuery(this).find('.et-pb-icon').text() === 'P') { jQuery(this).html('O'); // Change to minus } else { jQuery(this).html('P'); // Change back to plus } });

Solution

Data Pilot revolutionized Levi’s quality control process by developing an AI-powered computer vision platform. Utilizing the PyTorch framework, we trained a robust model to detect defects in garments. This model, integrated with Levi’s QUONDA application, enables real-time defect detection on iOS devices, thanks to TensorFlow Lite quantization. This innovative solution streamlines operations, improves product quality, and enhances customer satisfaction.

  • Building the Defect Detection Model
  • Integration Into the Client’s Application
  • Training the Model
/* By default all text modules are hidden */ .text-module { display: none; } /* Make the active text visible */ .text-module.active { display: block; } /* Default styles for list items */ .custom-timline-ul li { cursor: pointer; color:#000; } /* When a list item is active */ /* When a list item is active */ .custom-timline-ul li.active { color: #2271D8 !important; /* Change the text color to #2271D8 */ } /* Change the bullet color as well */ .custom-timline-ul li::marker{ color:#bebebe; } .custom-timline-ul li.active::marker{ color: #2271D8 !important; } .custom-timline-ul:before { content: ''; position: absolute; width: 2px; height: 30px; background: #F3F8FF; top: 0; left: -2px; }.custom-timline-ul:after { content: ''; position: absolute; width: 2px; height: 36px; background: #f3f8ff; bottom: 0; left: -2px; } @media (max-width:767px){ .custom-timline-ul:after { height: 41px !important; } .custom-timline-ul:before { height: 22px !important; } } @media (min-width:768px) and (max-width:1200px) { .custom-timline-ul:after { height: 36px !important; } .custom-timline-ul:before{ height:27px !important; } } @media (min-width:1401px){ .custom-timline-ul:after { height:46px !important; } } document.addEventListener("DOMContentLoaded", function () { console.log("DOM fully loaded and parsed"); // On page load, show text1 and highlight li1 showText(1); // Show the first text on load // Add click event to all list items let listItems = document.querySelectorAll('ul.main li'); console.log("List Items:", listItems); listItems.forEach(function (li, index) { console.log("Setting click event for li index:", index); li.addEventListener('click', function () { console.log("Clicked on li index:", index + 1); showText(index + 1); // Show corresponding text on click }); }); }); function showText(index) { console.log("Inside showText function, index:", index); // Hide all text modules let texts = document.querySelectorAll('.text-module'); console.log("Text modules:", texts); texts.forEach(function (text) { console.log("Removing active class from text module:", text); text.classList.remove('active'); }); // Show the selected text let selectedText = document.getElementById('text' + index); console.log("Selected text module:", selectedText); if (selectedText) { selectedText.classList.add('active'); } else { console.error("Text module not found for index:", index); } // Remove active class from all li elements let listItems = document.querySelectorAll('ul.main li'); console.log("Removing active class from all list items:", listItems); listItems.forEach(function (li) { console.log("Removing active class from li:", li); li.classList.remove('active'); }); // Add active class to the selected li let selectedLi = document.querySelector('ul.main li:nth-child(' + index + ')'); console.log("Selected li element:", selectedLi); if (selectedLi) { selectedLi.classList.add('active'); } else { console.error("List item not found for index:", index); } }

Data Pilot developed and deployed a computer vision model to automate the detection of defects in their garments. This model was integrated into the client’s quality control processes, significantly reducing the need for the client to manually intervene and improving its accuracy of catering to defects in their clothing. 

The AI model was integrated with the client’s existing QUONDA application, allowing real-time defect detection directly within their established workflows. This integration streamlined the quality assurance process and facilitated the client’s decision-making procedures. 

By utilizing Amazon SageMaker, Data Pilot implemented a robust pipeline for the initial training of the AI model. This involved processing the transformed data and training the computer vision model using the provided defected and non-defected images. 

The model was equipped to carry out the following for the client:

PyTorch Framework

PyTorch Framework was used as the primary deep learning library for developing the AI-powered object detection model.

TensorFlow Lite

The use of TensorFlow Lite ensured that the AI solution could be efficiently deployed on iOS applications, effectively extending the platform’s usability across mobile devices. 

The Impact

Reduced time spent on auditing garments

The AI model streamlined the defect detection process, reducing audit time significantly. Additionally, it minimized the number of manual clicks required by auditors, effectively decreasing human intervention and error.

High accuracy rate

The AI model achieved a successful accuracy rate. Because of good data quality, it proved its effectiveness during its Proof of Concept (POC) phase conducted across 2-3 factories.

 

Reduced defect logging time

The model reduced defect logging time significantly leading to a boost in productivity and reduced inspection times. This led to faster production cycles and quicker time-to-market for new clothing collections

 

Industry Applications

Quality Inspection in Manufacturing

In the manufacturing industry, AI-driven quality inspection solutions can be used to detect defects and inconsistencies in products during the production process. By integrating computer vision models, manufacturers can automate defect detection for various items such as automotive parts, electronics, or textiles. This helps in identifying faulty components early, reducing waste, minimizing recalls, and maintaining high-quality standards.  

Industry Applications

Quality Control in Retail

In the retail industry, such solutions can be applied to streamline quality control for merchandise before it reaches the end user. Retailers dealing with clothing, footwear, or accessories can integrate computer vision models into their supply chain to automatically inspect products for defects such as fabric tears, stitching errors, or incorrect labeling. By utilizing an AI-powered inspection process, retailers can significantly reduce returns and improve customer satisfaction by ensuring that only high-quality products reach the shelves.a 

“The workflow is very effective since we use DevOps to design and explain.”

Imran Saeed

Head of Customer Success, Triple Tree Solutions

Ready for your transformation? Let’s connect

Your transformation starts here.

Book Free Consultation

Speak with
our team
today!

hbspt.forms.create({ portalId: "21283101", formId: "24bebd07-23be-47a5-9214-0c941b5101f8" });