Vision AI, the technology that allows computers to “see” and understand the physical world, has been a mainstay in manufacturing for decades. From robotic arms picking up parts to automated optical inspection lines, cameras have played a crucial role in streamlining processes and ensuring quality. However, in recent years, the rise of artificial intelligence (AI) has injected a potent new wave into camera vision, transforming its capabilities and impact on the manufacturing landscape from niche automation use cases to a diverse solution that allows you to go deep and wide, and modernize across various disciplines.
Vision AI can also be an essential tool to improve revenue, manage costs, and, most critically, address workforce shortages in manufacturing. 72% of manufacturers view workplace acquisition as the number one challenge. In December of 2023, there were 601,000 unfilled manufacturing job openings, and 4 million more manufacturing jobs are expected to be required by 2030. Manufacturers struggle to attract and retain workers, and a lack of workforce will make it challenging to staff production lines, support expansion, and meet long-term strategic growth objectives. So how will your company mitigate these risks and tackle its future?
Whatever your company’s motivation, the key is to understand where to start, how to bring early time to value, and how best to scale and align to the many business challenges and opportunities associated with the application of vision AI.
What Has Changed?
While the technology, quality, and breadth of vision models have accelerated, the most notable change is eliminating the barrier to entry. Manufacturers of any size may benefit from these technologies due to lower infrastructure costs, accessible open-source models, and technological advances that make them realistic for any business.
- Tagging and Camera Vision AI: The ability to automatically tag and categorize objects in images, coupled with AI algorithms that can learn and adapt to variations, has vastly expanded the scope of what cameras can do. This has led to systems that identify objects to understand their relationships, contextualization, and may even predict future behavior.
- Lower Cost Cameras: The proliferation of high-resolution, affordable cameras has made deploying vision systems across diverse manufacturing environments more feasible. This democratizes the technology and allows smaller players to leverage its benefits. Some camera partners provide out-of-the-box vision AI functionality (aka, license plate detection or people detection), but these same cameras can be extended with advanced vision AI models. Whether your business requires low-cost IP cameras, high-end 8K cameras, or specialty cameras with high-speed industrial functionalities, the methodology for integrating any cameras with vision AI remains the same.
- Advancements in GPUs and Edge Compute: Powerful graphics processing units (GPUs) and edge computing technologies bring real-time processing capabilities closer to the data source, enabling faster responses and minimizing latency. This allows immediate inferencing and decision-making based on unstructured data, further optimizing production lines.
- Hyperscalers: Today’s hyperscalers are building vision AI into their platforms, removing barriers to this technology. These platforms integrate natively into their respective cloud stacks, integrate governance and security, and provide lifecycle support allowing organizations to label, tune, inference, and deploy. Depending on your business model and needs, many hyperscalers allow for inferencing in the cloud and deployment at the edge. Whichever you require, these solutions streamline MLOps and make it easier to manage over the life of the models, including deployment to edge compute.
What Are the Top Use Cases for Vision AI?
There aren’t many things you can’t accomplish given the right sample size, time, and money. With the advent of existing vision AI models detecting a wide range of objects, locations, text, and sentiments, it’s possible to detect and capture unstructured insight from images and videos to suit just about any use case. The good news is you don’t have to start with the most complicated use cases to get a business success. The following highlights some of the most feasible and widely adopted use cases in manufacturing that will get your organization up and running with meaningful business impact.
- Real-time Quality Control: High-resolution cameras combined with AI can inspect products at every production stage and identify minute defects that might escape the human eye. This leads to higher quality output, reduced waste, and improved brand reputation. For example, organizations that deployed AI-enabled quality control could see a reduction in defect rates by as much as 50% and save up to 50% on inspection costs. Given its high feasibility and business outcome, quality control is also the most widely adopted use case in manufacturing. If you want to start somewhere—start with quality control use cases.
- Predictive Maintenance: Cameras can now monitor equipment for subtle changes in vibration, temperature, movement, or wear patterns. AI algorithms analyze this data to predict potential failures before they occur, preventing costly downtime and ensuring smooth operation. Factory equipment often contributes to costly downtime, and implementing predictive maintenance can reduce unplanned maintenance events by as much as 30%.
- Optimized Production Lines: AI can analyze the flow of materials and products on a production line in real time, identifying bottlenecks and suggesting adjustments. This dynamic optimization increases throughput, reduces waste, and improves efficiency. A study found that AI-powered production line optimization can increase output by up to 25% and decrease waste by 15%. This is an excellent example of the evolution of machine vision towards a more holistic capability. Instead of just monitoring a single cell, cameras installed in the ceiling or over production lines now can reference a wide range of situations, allowing organizations to use cameras to gather more systemic insight within the plant and thus transform how they optimize their value stream.
- Optimized Robotics and Automation: AI vision enhances the capabilities of robots, equipping them with the ability to navigate complex environments, manipulate delicate objects, and adapt to changes in production workflows. This opens doors to increased automation, improved worker safety, and greater production flexibility. A study by the International Federation of Robotics found that collaborative robots assisted by vision systems can boost productivity by up to 40%.
- Enhanced Worker Safety: Most manufacturers track worker safety incidents after they’ve occurred. Cameras not only help determine the root cause for a particular incident, but they can also be used to identify unreported safety incidents, unreported injuries, and utilized to support a more proactive workplace safety culture. AI vision can detect unsafe practices and hazardous conditions in real-time, triggering alerts and even shut down equipment to prevent accidents. This can significantly improve worker safety and reduce workplace injuries, and ultimately drive the organization to be more proactive about its actions to protect employees.
Start Seeing Differently
Vision AI is not just a trend but a transformative force reshaping the manufacturing landscape. There is massive potential to collect unstructured data and translate that into valuable insight. Just imagine, whatever a supervisor, technician, or operator can see can often be emulated with a vision AI solution. This enables organizations to augment workforce gaps, eliminate mundane, error-prone, or non-value-added activities, and allow the workers we do have to perform value-added and thoughtful activities.
The power of vision AI to improve efficiency, quality, and safety is undeniable, and the collective advancements in this technology are paving the way for a more intelligent, productive, and competitive future for the manufacturing sector. As the technology evolves and becomes more accessible, expect to see even more innovation and disruption with expanding use cases, off-the-shelf foundational models, and integration with other AI capabilities like generative AI.
The key is not to wait. Most manufacturers need more skillsets and a broad understanding of applying this powerful technology in their business processes and technology stack. Starting early provides the ability to bring many stakeholders together, solve meaningful business challenges with realistic returns on investment, and prepare the organization for what will quickly follow. At the end of the day, those who start and execute with vision AI will outcompete those who don’t.
Our Manufacturing Practice has a team of experts from trade, an evolving portfolio of manufacturing solutions, and assists IT and OT teams by augmenting their existing skills with complimentary advisory services to help your business accelerate technology adoption where it matters most.
If your business wants to learn more about how we support our clients with artificial intelligence, engage our Helix Center for Applied AI and Robotics to learn more about this technology, available services, and the many use cases that may benefit your organization.
Ryan Spurr
Ryan Spurr is the Director of Manufacturing Strategy at Connection with 20+ years of experience in manufacturing, information technology, and portfolio leadership. He leads the Connection Manufacturing Practice, go-to-market strategy, client engagement, and advisory services focusing on operational technology (OT) and information technology that make manufacturers more digitally excellent.
