How CorePal uses GenAI to potentially recycle 98% of the valuable parts from junkyards to foster a circular economy
With today's increasing focus on sustainability, it is imperative to foster a healthy circular economy for a vertical industry by reducing waste. The key challenge, however, lies in the ability to identify parts when the data chain breaks towards the end of the product life cycle.
For automobiles, this means identifying valuable powertrain components ('cores') in messy junkyards, which would otherwise be scrapped 98% of the time, leading to a supply gap for remanufacturers on the reverse logistics side.
In collaboration with our partner in the core business, Vizuro developed CorePal, an AI agent platform that identifies thousands of core models without needing massive training data, to foster a circular economy that can potentially save 2 billion tons of carbon emission per year from recycling 75 million cores worldwide.
The Challenge: Finding Hidden Gems in the Junk
Every year, more than 30 million vehicles worldwide reach the end of their useful life. While many components are beyond repair, others – particularly powertrain cores – still hold significant value for remanufacturing – on average $1,500 increase in resale value, plus 3 tons of reduced carbon emission per core. The challenge? Accurately identifying and sourcing these valuable parts from junkyards on a larger scale.
Traditionally, this process relied on specialized workers known as "Core Guys", who would drive around multiple junkyards daily to scour suitable cores. This approach, while functional, created several inefficiencies:
Lack of scalability: core sourcing is heavily limited by Core Guys’ knowledge and training through apprenticeship, as well as its labor intensiveness.
Inconsistent supply quality: discrepancies between model information provided by Core Guy and actual cores received by remanufacturers.
Lack of market visibility: no online marketplace for supply and demand of cores, causing fragmented, inefficient local transactions like ride hauling before Uber.
The Solution: CorePal's One-Shot Training Using GenAI
Vizuro's vision to address these challenges led to CorePal – an online marketplace for automotive parts identified and validated by machine vision AI. However, one major hurdle is the prohibitive cost to collect massive training images for thousands of specialized parts. Our solution? Rather than relying on extensive libraries of 2D images, CorePal took a road less taken:
One-shot data collection: using 3D scanners to capture detailed digital twins of the target parts.
Synthetic data generation: using GenAI to render these digital twins into thousands of synthetic 2D images to mimic various real-world conditions.
Adaptive AI training: applying the rich dataset to train accurate image classifiers, boosting accuracies with additional synthetic images where errors tend to occur.
Agentic platform deployment: packaging the solution into a multitier platform – mobile app, scanning stand, robotic AGVs, and surveillance drones.
Sounds familiar? This approach echoes with the recent Omniverse movement touted by Nvidia, only two years earlier.
The Edge: CorePal Advantages
What makes CorePal's approach unique is its deployment feasibility which can be easily generalized for other specialized products. Unlike common objects with abundant online images, specialized automotive components, such as Nissan JF010E or Toyota U140E, have extremely limited training data available. CorePal makes sure any product, no matter how specialized, can enjoy the following advantages:
Performance robustness: Each 3D scan of a component can generate tens of thousands of synthetic 2D images from various viewpoints, lighting, and backgrounds, etc., to mimic real-world conditions the AI may encounter.
Cost efficiency: using GPUs to generate synthetic images proves significantly more cost-effective compared to manual image acquisition.
Rapid deployment: CorePal allows pretraining even before the product hits production as long as its digital twins (CAD files or 3D scans) are available, enabling rapid deployment of reverse logistics in any product lifecycle planning.
The Results: Efficiency, Accuracy, and Cost Savings
After the initial pilot test in 2022 to successfully identify 50 automotive transmission models, CorePal was quickly scaled up to classify 500 transmission and torque converters with high accuracies, ready for commercial deployment as a mobile app. During 2024, CorePal was further integrated with a scan hardware platform and deployed across multiple automotive recycling facilities. Overall, CorePal has yielded the following outcomes:
92-94% accuracy in identifying complex automotive cores, such as engines and transmissions
47% reduction in inventory management time through automated identification and logging
200~300% productivity boost for human workers with a single CorePal device
Streamlined operations through digital inventory management and marketplace transaction support
Beyond Recycling: Creating a Sustainable Future
CorePal's impact extends far beyond immediate operational efficiencies:
Carbon reduction: by facilitating remanufacturing, CorePal helps reduce waste and carbon emissions by 66% compared to making new cores.
Resource conservation: while preventing valuable components from being scrapped, the supply gap in automotive cores is also bridged to foster a circular economy.
Data-driven insights: with the circular economy cranking, the supply-demand data in real time also provide rich analytics to enhance market efficiencies, such as sourcing and pricing.
Cross-industry applications: this versatile technology platform can easily adapt to any other verticals, such as electronic waste management and warehouse operations.
Key Takeaway: Technologies Can Drive Circular Economy and Sustainability
CorePal demonstrates how cutting-edge AI technologies can transform a vertical industry to be more environmentally friendly and sustainable. By simply empowering stakeholders with object identification capabilities to augment the broken data chain, the circular economy for automobiles can be rejuvenated to be more efficient and viable. In addition, as industries worldwide face common pressure for sustainability, CorePal shows that environmental responsibility and the economy can indeed go hand-in-hand.
Ready to Transform Your Recycling Operations?
Discover how Vizuro’s AI-powered solution like CorePal can strengthen your data chain to foster a circular economy, reduce labor costs, and achieve your sustainability goals.
About CorePal
CorePal is an advanced AI-driven remanufacturing solution developed by Vizuro in collaboration with Berkeley Standard. By leveraging multi-modal AI technology, CorePal streamlines the identification, sourcing, and restoration of automotive core components, driving efficiency and sustainability in the automotive recycling industry.
Designed for adaptability, CorePal’s technology extends beyond automotive applications, offering innovative solutions for industrial component tracking and electronic waste management.
Learn more about Vizuro solutions: https://www.vizuro.com/solution
Reach out to Vizuro team: https://www.vizuro.com/contact
Comments