Demystifying Image Recognition for Shape-Shifting Objects: The Power of ZIA Capture

A 3D digital twin of a Doritos crisp packet, demonstrating deformability for improved IR training.

In the dynamic realm of consumer packaged goods (CPG), managing the intricacies of soft body packaging deformations presents a formidable challenge for brands.

Many Image Recognition solutions struggle to detect soft products, making catalogue onboarding problematic for brands and field agents. 

Neurolabs' innovative solution, ZIA, optimises catalogue onboarding processes and offers a transformative approach to image recognition in the CPG industry.

Using synthetic data and computer vision, Neurolabs is changing how CPG brands onboard products, making the process more efficient, cost-effective, and reliable. 

In this blog, we’ll explore how ZIA and ZIA Capture work together to make product onboarding simple and effective, even for the most complex CPG products. 

Understanding Product Deformations

Shelves in a grocery store showcasing 'deformed' products (crisp packets).

Soft body packaging materials, characterised by their flexibility and adaptability, are prone to deformations that can significantly impact product recognition. 

Stand-up pouches with zip locks, vacuum pouches, and laminated tubes are just a few examples of packaging susceptible to deformations.

However, even rigid packaging, such as aerosol spray cans and beverage bottles, is not immune to dents and distortions. 

Image Recognition & The Challenges Caused by Deformations

Deformations pose multifaceted challenges for CPG brands, ranging from stock management discrepancies to compromised shelf visibility and product recognition technology. 

The above illustrates how programmatically manipulating synthetically generated data (3D models) enhances our image recognition's ability to accurately identify deformable product packaging in real-world scenarios.

The reliance on real data adds another layer of complexity to this challenge. Algorithms trained on a subset of real deformations may struggle to understand deformations outside their training set, limiting their ability to recognise deformed products accurately.

The reliance on real data also means that algorithms may only understand deformations they have been trained on, which is a subset of the potential deformations that can occur in real-world scenarios. This limitation hampers the algorithm's generalisation ability for any possible deformation, hindering accurate product recognition.

Deformations on products are typical, especially with soft-body SKUs. For example, picking up a packet of crisps from a shelf can deform it in numerous ways, such as wrinkles and folds. This can make it visually unrecognisable to image recognition (IR) software when it’s placed back onto the shelf. 

Ensuring accurate data collection and insights is paramount for informed business decisions and strategy development. Without precise product detection capabilities, brands risk misinterpreting market trends, impacting product ranges, and compromising overall business performance. 

However, relying solely on extensive data collection efforts is time-consuming, costly, and prone to human error. CPG brands need a new, effective technological solution, and that’s where Neurolabs comes in. Our innovative approach leverages advanced machine learning techniques to overcome the limitations of traditional data reliance, enabling accurate and efficient product recognition even in the face of diverse and unpredictable deformations.

The Benefits of Neuorlabs’ Synthetic Data Approach 

Synthetic data emerges as a game-changing solution, offering distinct advantages over real-world data for image recognition applications. 

Neurolabs' digital twin approach represents a groundbreaking innovation in the realm of synthetic data-driven technology. Our digital twins are meticulously crafted 3D representations of real-world objects, generated using advanced computer vision techniques and proprietary algorithms. 

Using artwork or images captured via our app (ZIA Capture) to create a 3D model, these digital replicas faithfully capture the intricate details and characteristics of physical objects, enabling unparalleled accuracy in image recognition tasks. 

By programmatically generating synthetic data and 3D models, Neurolabs bypasses the limitations of traditional data collection methods. This enables data representing all types of deformations to be produced at scale and allows the algorithm to train with more diversity and faster than real data variations could ever provide.

Synthetic data enables the simulation of realistic deformations across various packaging materials, enabling our technology,  ZIA, to train algorithms more effectively than ever before. This approach ensures enhanced flexibility, scalability, and accuracy, revolutionising image recognition processes in the CPG landscape.

Pioneering Automated Deformation

Neurolabs' ZIA stands at the forefront of automated geometric deformation technology, pioneering accurate product detection in retail environments. By simulating deformities using synthetic computer vision, ZIA delivers unparalleled accuracy, with recognition rates surpassing 95% across various product categories.

Deployed in real-world scenarios, ZIA's cutting-edge capabilities empower brands to streamline shelf auditing processes and mitigate product and human detection errors. Integrating ZIA into existing solutions or apps enhances efficiency and effectiveness, offering a seamless transition to advanced image recognition technology.

The ZIA Advantage: Faster & More Efficient Shelf Audits

ZIA revolutionises the catalogue onboarding and management process, offering a faster and more efficient solution for CPG brands. With ZIA, brands can decrease onboarding time, eliminate human errors, and enhance visibility across their product catalogues.

ZIA's 'Universal Artwork Approach' empowers users to effortlessly create digital twins from existing PDF and PNG artwork files of manufacturing labels, ensuring accurate and easy product onboarding. By leveraging a CPG's master DAM (Data Asset Management) catalogue or directly integrating with third-party DAM providers, you can easily upload or find product artwork. 

The ZIA Capture app streamlines the process of obtaining artwork for products without it. Field agents or internal officers can use the app to quickly and accurately capture product images, creating a 3D model for your catalogue in mere minutes.

By leveraging synthetic data and advanced computer vision, ZIA empowers brands to overcome deformation-related challenges, achieve unprecedented accuracy, and unlock new opportunities for success.

Additionally, by seamlessly integrating with existing systems and platforms, ZIA and ZIA Capture facilitate collaboration between field agents and catalogue managers, driving growth and innovation in the CPG industry.

The Future of In-Store Execution with ZIA Capture

As the CPG landscape evolves, the need for innovative solutions to address deformation-related challenges becomes increasingly apparent. ZIA emerges as a transformative tool, empowering brands to navigate the complexities of soft body packaging deformations with confidence and precision.

With ZIA Capture, the future of in-store execution is brighter than ever before. By embracing synthetic data-driven solutions and harnessing the power of advanced image recognition technology, brands can unlock new levels of efficiency, accuracy, and growth in the dynamic consumer packaged goods world.

Join us in shaping the future of retail execution with ZIA. Schedule a demo today and discover how Neurolabs revolutionises catalogue onboarding for CPG brands worldwide. 

---

At Neurolabs, we are revolutionising in-store retail performance with our advanced image recognition technology, ZIA. Our cutting-edge technology enables retailers, field marketing agencies and CPG brands to optimise store execution, enhance the customer experience, and boost revenue as we build the most comprehensive 3D asset library for product recognition in the CPG industry.