- Anticipate what’s next — and act with confidence
- Success stories
- Sport Group Implements AI to Enhance Quality Standards
Sport Group Implements AI to Enhance Quality Standards
98% detection accuracy
Only 2% of defects escape automated inspection.
Only 3% manual checks
Just 3 rolls in every 100 require human quality control.
20 weeks to production
Full AI solution implemented and live in under 5 months.
Challenge
Sport Group leads the world in the manufacture and supply of high-performance sports surfaces. The group sells and installs more synthetic turf sports fields, athletic tracks and courts than any other business. With a product portfolio that comprises many famous brands, including AstroTurf, Polytan and SynLawn, the group is united in a commitment to innovation and excellence.
In addition to offering world-class performance and durability, sports surfaces must look perfect. Sport Group surfaces are constantly exposed to the intense scrutiny of players and spectators in global locations. Cosmetic defects, such as stripes, curling or variations in colour or tone are simply unacceptable and would reflect badly on a brand and the group. Sport Group strives for perfection in both performance and appearance. But some visual defects do not appear until the surface is complete or about to be installed. Thorough quality inspection is a critical part of the group’s brand image.
As part of its commitment to innovation, Sport Group wished to integrate Visual Inspection into its manufacturing and production process. If successful, an autonomous solution would reduce the need for manual inspection, saving time and money while maintaining consistent product quality.
Having delivered several successful visual inspection projects, GFT was the ideal partner for Sport Group and was asked to lead the project. The assignment included all aspects of Visual Inspection, from hardware/camera selection and procurement to implementing and tuning Vertex AI models. Sport Group was confident in the technology and decided to skip the proof-of- concept stage in favour of a preliminary trial in a live production environment.
“As Premium Google Cloud Partner of the Year in Germany, with a strong track record of success using Visual Inspection, GFT was ideally positioned to support Sport Group. We were happy to have a trusted partner who could manage all aspects of this complex project.”
Engagement
GFT assembled a bespoke multidisciplinary project team that included data scientists, cloud engineers and hardware experts. At the outset GFT documented the scope of the project in a detailed plan that included project goals and milestones.
The aim of the project was to prove Visual Inspection capabilities at scale in a live environment. It was envisaged that implementation would take around eight weeks excluding hardware procurement.
With cameras and hardware in place GFT managed the testing and fine-tuning of the hardware setup remotely. Other highlights of the project included:
- Development of a data pipeline for image sources including pre-processing via filters and optical character recognition (OCR) technology
- Training of Vertex AI models with live data to identify defects according to 12 colour criteria
- Automatic upload from the local hard disk to the Google Cloud pipeline
- Collation and Reporting of Visual Inspection and OCR results to produce usable metrics
- Stable operation of pipelines and AI models and thorough testing with real production images
- Safe archiving of images and data in GCP cloud storage
From the outset, the trial highlighted production defects early. During the trial, the main challenge was to reduce the number of false positives and to pinpoint actual defects accurately and consistently. This challenge was met by improving data quality and further training of the model.
The model consistently delivered results within acceptable tolerances of a two-percent escape rate meaning that almost all defects are detected automatically. And a three-percent overkill rate meant that only three rolls in every 100 require manual quality inspection.
As the project progressed results became more accurate and AI models were refined. Sport Group soon realised how Visual Inspection could boost quality inspection, save substantial time and money.
“Sport Group is committed to growth through innovation, and we treated Visual Inspection as a strategic initiative. Through our strong partnership with GFT, we could test the new technology in a production environment at scale and deliver benefits confidently and quickly.”
“This solution has exceeded our expectations. With an overkill rate of around three percent, only two rolls of turf in every 100 must be checked manually. And with better images and model refinement, this number will reduce even more.”
consulting, and innovative solutions for businesses worldwide. Explore GFT's services in AI, cloud
computing, and technology consulting.
Benefit
From project setup to implementation, the Sport Group Visual Inspection integration took around 20 weeks. The trial proved that the new quality inspection solution could detect defects very quickly and accurately. The solution takes Sport Group nearer to its goal of perfection.
The Sport Group vision was to integrate Visual Inspection into all its production facilities and the trial proved that Visual Inspection could be implemented quickly and with no interruption to business as usual. Once the hardware is configured, GFT can complete the setup remotely, including data management and training of models. The GFT aim is to empower Sport Group to become self-sufficient as soon as practicable.
Now every production run undergoes a Vertex AI defect detection process, according to 12 colour criteria and every batch has a detailed quality assurance support. The model has consistently delivered results within acceptable tolerances of a two percent escape rate and three percent overkill rate. As well as detecting defects, the new solution categorises them according to type. New evaluation criteria can be included as the Visual Inspection model becomes more sophisticated.
Sport Group is confident that every batch of turf produced is accompanied by a quality assurance report which demonstrates the groups commitment to perfection. Model performance can be further increased through additional images and labelling.
Although the model runs autonomously, GFT continues to refine the solution to improve reporting of results, for example using dashboards. Sport Group plans to roll out Visual Inspection in global locations, including Australia and the USA, and is also defining further use cases.
“Visual Inspection has truly delivered on its promise to transform quality inspection. We at Sport Group are committed to rolling out this technology at all its production facilities worldwide as we welcome a new era of automated quality assurance.”
Download - just one click away
-
Success story
Sport Group Adopts AI to Enhance Quality Standards
