Why did humans evolve the eyes we have today?While scientists can’t go back in time to study the environmental pressures that shaped the evolution of the diverse vision systems that exist in nature, a new computational framework developed by MIT researchers allows them to explore this evolution in artificial intelligence agents.The framework they developed, in which embodied AI agents evolve eyes and learn to see over many generations, is like a “scientific sandbox” that allows researchers to recreate different evolutionary trees. The user does this by changing the structure of the world and the tasks AI agents complete, such as finding food or telling objects apart.This allows them to study why one animal may have evolved simple, light-sensitive patches as eyes, while another has complex, camera-type eyes.The researchers’ experiments with this framework showcase how tasks drove eye evolution in the agents. For instance, they found that navigation tasks often led to the evolution of compound eyes with many individual units, like the eyes of insects and crustaceans.On the other hand, if agents focused on object discrimination, they were more likely to evolve camera-type eyes with irises and retinas.This framework could enable scientists to probe “what-if” questions about vision systems that are difficult to study experimentally. It could also guide the design of novel sensors and cameras for robots, drones, and wearable devices that balance performance with real-world constraints like energy efficiency and manufacturability.“While we can never go back and figure out every detail of how evolution took place, in this work we’ve created an environment where we can, in a sense, recreate evolution and probe the environment in all these different ways. This method of doing science opens to the door to a lot of possibilities,” says Kushagra Tiwary, a graduate student at the MIT Media Lab and co-lead author of a paper on this research.He is joined on the paper by co-lead author and fellow graduate student Aaron Young; graduate student Tzofi Klinghoffer; former postdoc Akshat Dave, who is now an assistant professor at Stony Brook University; Tomaso Poggio, the Eugene McDermott Professor in the Department of Brain and Cognitive Sciences, an investigator in the McGovern Institute, and co-director of the Center for Brains, Minds, and Machines; co-senior authors Brian Cheung, a postdoc in the  Center for Brains, Minds, and Machines and an incoming assistant professor at the University of California San Francisco; and Ramesh Raskar, associate professor of media arts and sciences and leader of the Camera Culture Group at MIT; as well as others at Rice University and Lund University. The research <a href="https://doi.org/10.1126/sciadv.ady2888" target="_blank">appears today in Science Advances</a>.<h2>Building a scientific sandbox</h2>The paper began as a conversation among the researchers about discovering new vision systems that could be useful in different fields, like robotics. To test their “what-if” questions, the researchers decided to <a href="https://mit-genai.pubpub.org/pub/bcfcb6lu/release/3" target="_blank">use AI to explore the many evolutionary possibilities</a>.“What-if questions inspired me when I was growing up to study science. With AI, we have a unique opportunity to create these embodied agents that allow us to ask the kinds of questions that would usually be impossible to answer,” Tiwary says.To build this evolutionary sandbox, the researchers took all the elements of a camera, like the sensors, lenses, apertures, and processors, and converted them into parameters that an embodied AI agent could learn.They used those building blocks as the starting point for an algorithmic learning mechanism an agent would use as it evolved eyes over time.“We couldn’t simulate the entire universe atom-by-atom. It was challenging to determine which ingredients we needed, which ingredients we didn’t need, and how to allocate resources over those different elements,” Cheung says.In their framework, this evolutionary algorithm can choose which elements to evolve based on the constraints of the environment and the task of the agent.Each environment has a single task, such as navigation, food identification, or prey tracking, designed to mimic real visual tasks animals must overcome to survive. The agents start with a single photoreceptor that looks out at the world and an associated neural network model that processes visual information.Then, over each agent’s lifetime, it is trained using reinforcement learning, a trial-and-error technique where the agent is rewarded for accomplishing the goal of its task. The environment also incorporates constraints, like a certain number of pixels for an agent’s visual sensors.“These constraints drive the design process, the same way we have physical constraints in our world, like the physics of light, that have driven the design of our own eyes,” Tiwary says.Over many generations, agents evolve different elements of vision systems that maximize rewards.Their framework uses a genetic encoding mechanism to computationally mimic evolution, where individual genes mutate to control an agent’s development.For instance, morphological genes capture how the agent views the environment and control eye placement; optical genes determine how the eye interacts with light and dictate the number of photoreceptors; and neural genes control the learning capacity of the agents.<h2>Testing hypotheses</h2>When the researchers set up experiments in this framework, they found that tasks had a major influence on the vision systems the agents evolved.For instance, agents that were focused on navigation tasks developed eyes designed to maximize spatial awareness through low-resolution sensing, while agents tasked with detecting objects developed eyes focused more on frontal acuity, rather than peripheral vision.Another experiment indicated that a bigger brain isn’t always better when it comes to processing visual information. Only so much visual information can go into the system at a time, based on physical constraints like the number of photoreceptors in the eyes.“At some point a bigger brain doesn’t help the agents at all, and in nature that would be a waste of resources,” Cheung says.In the future, the researchers want to use this simulator to explore the best vision systems for specific applications, which could help scientists develop task-specific sensors and cameras. They also want to integrate LLMs into their framework to make it easier for users to ask “what-if” questions and study additional possibilities.“There’s a real benefit that comes from asking questions in a more imaginative way. I hope this inspires others to create larger frameworks, where instead of focusing on narrow questions that cover a specific area, they are looking to answer questions with a much wider scope,” Cheung says.This work was supported, in part, by the Center for Brains, Minds, and Machines and the Defense Advanced Research Projects Agency (DARPA) Mathematics for the Discovery of Algorithms and Architectures (DIAL) program.

Researchers developed a computational framework that enables them to explore and probe the evolution of vision systems over millions of years using embodied AI agents. Image: iStock

The AI-powered tool could inform the design of better sensors and cameras for robots or autonomous vehicles.

A "scientific sandbox" lets researchers explore the evolution of vision systems

Matroid builds no-code computer-vision detectors that can spot everything from microscopic material defects to real-time safety hazards on a factory floor. But “no-code” doesn’t mean “no humans.” Every model we deploy is the result of seamless teamwork between our engineers—who design the algorithms—and our data annotation specialists—who teach those algorithms what the world actually looks like. This series, The Visionary Mind, pulls back the curtain on that partnership so readers can see how human insight and machine precision combine to push computer vision forward.When people think about AI that can “see,” whether it is a self-driving car, a medical imaging system, or intelligent surveillance, they usually picture the end product. What they do not see is the massive amount of human effort behind the scenes. Data annotators are the ones who carefully label images, videos, and everything in between so that machines can actually learn. Think of annotation as digital apprenticeship: every bounding box, polygon, or class label is a tiny lesson we hand to the algorithm, saying, “This is what matters—and here’s why. Data annotation is not only about drawing boxes. Some days, it means painstakingly outlining tiny details pixel by pixel. Other days, it means making judgment calls in messy real-world scenarios.Those judgment calls are where the human element shines. A streetlight obscured by fog, a surgical implant partially hidden by glare—these are puzzles that require intuition, context, and sometimes heated team debates. We’re not just labeling pixels; we’re encoding collective experience so future detectors can make life-or-death distinctions in milliseconds.One of the best parts of this work is seeing its real-world impact. Our team has contributed to projects across various industries, including aerospace, healthcare, and security. While we can’t publicly discuss the specifics, one project that stands out involved a client in the security sector. We labeled hundreds of thousands of images, including rare and challenging scenarios, training the AI not only to detect objects but also to understand context, such as identifying early signs of potentially dangerous situations. Knowing that our work could help prevent real-world incidents is what makes it worth it. For instance, picture a crowded transit hub at rush hour: a single unattended backpack appears beneath a stairwell, partially hidden by shadows and foot traffic. A typical camera feed would record the scene passively, but our trained detector flags the bag within seconds—distinguishing it from harmless luggage by cross-referencing behavior patterns (no accompanying person, stationary for too long) and environmental cues (proximity to restricted zones). Security personnel receive an alert before most commuters have even noticed the risk, giving them precious minutes to investigate and, if necessary, evacuate. Scenarios like this—entirely hypothetical yet technically feasible—show how context-aware annotation turns raw pixels into early warnings that can save lives.Moments like these remind us that annotation isn’t busywork—it’s an act of stewardship. Every false negative we eliminate could be a prevented accident; every edge case we anticipate could keep critical environments safer or a patient healthier.Of course, there are challenges. Visual data in the real world is messy. Objects are not always apparent. Lighting can be terrible. Sometimes, you are making split-second decisions about whether that blurry figure is a person, a signpost, or just a strange shadow. It not only requires precision but also demands a lot of critical thinking and situational awareness.Consistency is another non-negotiable part of the job. Small mistakes might not seem like a big deal at first, but even tiny inconsistencies can seriously confuse machine learning models. It is a role where attention to detail is not only appreciated but also essential, especially when working with datasets consisting of hundreds of thousands of images.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1764753659335-object-detection.webp" style="width:100%px" class="fr-fic fr-dib" />&lt;figure&gt;(Object Detection with Bounding Boxes)&lt;/figure&gt; <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1764753679834-segmented-annotation.webp" style="width:100%px" class="fr-fic fr-dib" />(Segmented Annotation)The field is also evolving quickly. New tools are constantly emerging, including semi-automated platforms that suggest annotations or pre-label parts of images. They make the work faster, but they also add a new layer of responsibility. We are not just labeling anymore. We are quality control for the machines, making sure their guesses are accurate and grounded in reality.Despite the challenges, it is an exciting place to be. Computer vision is expanding rapidly, and the applications are becoming more ambitious, from diagnosing diseases earlier to creating more innovative augmented reality experiences. While AI tends to get the spotlight, it is the careful, often invisible groundwork —the thousands of labeled frames —that makes it all possible. In other words, if AI is the orchestra, annotation is the sheet music. Remove it, and all you have left is noise.Looking ahead, I’m excited for what’s next. As the field continues to grow, so does the importance of the work behind it. No matter how advanced AI becomes, it still needs people to guide it, correct it, and teach it how to make sense of the world. Being part of that foundation is not just rewarding; it is a reminder that real progress always starts with the details.

Matroid builds no-code computer-vision detectors that can spot everything from microscopic material defects to real-time safety hazards on a factory floor.

A Day in the Life of a Data Annotator: Behind Computer Vision

Modern industrial automation works best when smart vision systems and robots team up. A great example of this is a project by Tiesse Robot for Fonderia di Torbole, a well-known Italian foundry that specializes in parts for the automotive industry.The aim was to fully automate the handling of brake discs, starting with picking the raw parts and ending with getting them ready for painting.The system includes three connected robotic stations that work together, relying on Photoneo’s 3D vision to guide the robots with accuracy and reliability – even in messy, unpredictable environments.<iframe width="640" height="360" src="https://www.youtube.com/embed/GIsJCtujBRQ?&amp;wmode=opaque&amp;rel=0&amp;enablejsapi=1&amp;origin=https://www.wevolver.com" frameborder="0">     </iframe><h3>Project overview:</h3><ul><li>End Customer: Fonderia di Torbole </li><li>Integrator: Tiesse Robot </li><li>Application: Robotic Bin Picking of Raw Brake Discs (Brake Discs Picking)</li><li>Photoneo Products: PhoXi 3D Scanner XL, Bin Picking Studio</li><li>Distributor: The whole project was made possible thanks to a strong collaboration with our local distributor, Basler Italy. </li></ul><img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzU5MTQ2NzQzMTQzLVRpZXNzZV9QaG90b25lbyAwMi5qcGciLCJlZGl0cyI6eyJyZXNpemUiOnsid2lkdGgiOjk1MCwiZml0IjoiY292ZXIifX19" style="width:100%px" class="fr-fic fr-dib" />Overview of the factory floor<h2>The Challenge: Automating the Unpredictable</h2>The goal was to fully automate the handling of heavy brake discs, from a semi-structured bin to the final preparation for painting. The initial stage, bin picking, presented a significant hurdle. The raw parts – cast iron and steel brake discs weighing up to 15 kg each – are piled in large bins. A successful automation solution needed to:<ul><li>Reliably identify and locate discs in any orientation within a large bin.</li><li>Handle heavy, metallic parts with potentially reflective or dark surfaces.</li><li>Provide precise 3D coordinates to a robot for a secure grip.</li><li>Calculate collision-free paths for the robot arm and gripper to prevent damage to the parts, the bin, or the robot itself.</li><li>Achieve a fast cycle time (under 45 seconds) to feed the rest of the production line without interruption.</li></ul><h2>The Solution: Intelligent 3D Vision by Photoneo</h2>Tiesse Robot built the first of three robotic islands around a <a href="https://www.photoneo.com/bin-picking-studio/kawasaki/" target="_blank">Kawasaki</a> RS080N robot guided by Photoneo’s cutting-edge bin picking solution. This combination transformed the challenging task into a reliable and efficient automated process.The system leverages two core Photoneo technologies:<h3>The Role of PhoXi 3D Scanner XL </h3>Chosen for its large scanning volume and high-resolution output, the <a href="https://ai.photoneo.com/best-3d-scanner" target="_blank">PhoXi 3D Scanner</a> is perfectly suited for large industrial bins. Using advanced structured light technology, it captures a detailed and accurate 3D point cloud of the entire bin in a single scan, effectively mapping the position and orientation of every visible brake disc, regardless of their chaotic arrangement. Its robust build ensures flawless performance in the demanding foundry environment.<h3>The Role of Bin Picking Studio </h3>This powerful vision intelligence software serves as the brain of the operation. After the PhoXi 3D Scanner XL captures the scene, Bin Picking Studio instantly processes the point cloud.<ul><li>Localization &amp; Part Recognition: The software identifies every pickable brake disc, determining its precise 3D position and orientation.</li><li>Path Planning &amp; Collision Avoidance: Crucially, it calculates the optimal, collision-free trajectory for the Kawasaki robot to approach, grip, and extract the selected disc from the cluttered bin.</li><li>Robot Communication: It seamlessly communicates the picking coordinates and path to the robot controller, enabling autonomous operation.</li></ul><h3>A Seamless, Vision-Guided Workflow</h3>In the first robotic island, the automated sequence relies on 3D vision for maximum efficiency:<ol><li>The PhoXi 3D Scanner XL performs a scan of the bulk bin containing the raw brake discs.</li><li>Bin Picking Studio analyzes the 3D data, identifies the best candidate for picking, and calculates a safe extraction path.</li><li>The software provides the orientation data to the Kawasaki robot, which uses a magnetic gripper to securely pick the disc.</li><li>If Bin Picking Studio indicates a disc is upside down, the robot automatically performs a flipping maneuver before placing the part correctly onto a powered roller conveyor, ensuring it is perfectly prepared for the next stage of turning.</li></ol>This flawlessly executed first step guarantees a continuous and synchronized flow of material to the subsequent islands, which handle post-turning operations, inspection, and final outfeed.<h2>How It Works: A Three-Stage Automated Process</h2>The entire solution is orchestrated across three interconnected robotic islands, each handling a specific phase of the production cycle to ensure a continuous and synchronized workflow from raw part to finished product.<h3>Stage 1: Vision-Guided Brake Discs Picking</h3>The process begins at the first island, where a Kawasaki Robotics RS080N robot is tasked with retrieving raw brake discs from bulk bins.Guided by a Photoneo PhoXi 3D Scanner, the robot identifies the precise orientation of each disc. Using a magnetic gripper, it securely picks a correctly positioned part – or flips an upside-down one – before placing it onto a powered roller conveyor that feeds the turning line. <h3>Stage 2: Post-Turning Operations </h3>After turning, the discs arrive at the second island for intermediate machining. Here, a high-payload Kawasaki Robotics ZX165U robot, equipped with a dual pneumatic three-jaw gripper, takes over. It first uses a 2D vision system (TS-Vision) to determine the correct rotational position (phasing) of the disc. The robot then manages the loading and unloading of two drilling units and a subsequent balancing station. Based on the results, compliant discs are sent to a washing machine, while any rejects are automatically diverted to a separate conveyor.<h3>Stage 3: Final Inspection and Outfeed </h3>In the final stage, the washed discs undergo inspection and marking. Another Kawasaki Robotics ZX165U robot handles the parts, moving them between several units. The disc first undergoes a dimensional inspection, followed by unique identification marking with a laser for full traceability. Finally, the robot places the finished disc either onto the customer’s outfeed roller conveyor or onto a transfer belt leading to the painting system, completing the automated cycle.<h2>Results: The Foundation for Flawless Brake Discs Picking</h2>By integrating Photoneo’s 3D vision, Tiesse Robot provided the foundry with a solution that unlocked complete automation for the brake disc line. The Photoneo-powered bin picking cell delivered:<ul><li>Unmatched Reliability: Consistent and accurate picking of heavy, randomly oriented parts, eliminating manual labor for a strenuous and repetitive task.</li><li>High Throughput: The system met the demanding cycle time of 45 seconds, ensuring the entire production line operates at peak efficiency.</li><li>Intelligent Automation: The combination of 3D scanning and advanced path-planning software provides the intelligence needed for a smooth operation.</li><li>Process Continuity: By solving the most complex part of the process first, Photoneo’s technology became the cornerstone of the entire automated system, enabling a seamless end-to-end workflow.</li></ul><hr />Curious how can PhoXi 3D Scanner solve your automation challenge? <a href="https://ai.photoneo.com/best-3d-scanner" target="_blank" rel="noopener noreferrer">Take a look at the latest generation of our revolutionary 3D scanner</a> for the most demanding industrial tasks.

High-Speed Automation for Automotive Parts Foundry

Case Study: Solving Heavy Brake Disc Picking

Imagine being an R&amp;D engineer at Precision Parts Inc., a fictional global manufacturer renowned for its intricate automotive components. Its Quality Assurance Manager – we’ll call her Sarah – comes to you with a significant challenge. The company’s current manual visual inspection process for finished parts is leading to inconsistent quality, slow defect detection, and a concerning amount of material waste due to human error and fatigue. Sarah knows she needs an automated solution that is both precise and efficient, and asks you to find a way to integrate smart technology into the production line. Where would you look?Luckily, you’ve heard about the <a href="https://store.arduino.cc/products/portenta-proto-kit-ve" target="_blank">Arduino Portenta Proto Kit VE</a> (Vision Environment) and are ready to test its comprehensive capabilities for rapid prototyping in machine vision and environmental monitoring.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1756885389929-AKX00074_04.front.jpg" style="width:100%px" class="fr-fic fr-dib" /><h3>Motion, Vision and Environmental Sensing: All in One Kit</h3>For the critical aspect of vision and edge AI, you position a <a href="https://store.arduino.cc/products/nicla-vision" target="_blank">Nicla Vision</a> module – a compact color camera designed for AI on the Edge, included in the kit – directly above the test rig, mounted on a robotic arm. This setup is able to continuously capture images of each component as it moves around the Device Under Test at a speed of 20 frames per second.To simulate real-world conditions, you create a test rig that mimics the actual inspection station: plastic parts mounted on a bigger component under the camera’s field of view. Leveraging the compact form factor of the Nicla, the core of this part of the project involved training a basic convolutional neural network (CNN), designed to identify and classify a range of surface defects. This CNN was specifically engineered to detect critical manufacturing anomalies, including precise mounting positioning, ensuring components are placed correctly and securely, and verifying the existence of all required components, Regarding control and actuation, if the AI model detects a defective component, the Nicla Vision is connected to a <a href="https://store.arduino.cc/products/portenta-h7" target="_blank">Portenta H7</a> through the <a href="https://store.arduino.cc/products/portenta-mid-carrier" target="_blank">Portenta Mid Carrier Proto Shield</a> (both included in the kit), instantly triggering a visual alarm. The alarm stops the arm indicating the faulty component. To provide an immediate alert to nearby operators, you also decide to incorporate a Buzzer Modulino node (also included in the kit!), which emits an audible signal whenever a defect is identified. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1756886114879-AKX00074_01.iso.jpg" style="width:100%px" class="fr-fic fr-dib" /><h3>From Compatibility to Easy Connections</h3>The Portenta Mid Carrier Proto Shield’s compatibility with Arduino UNO shields, Qwiic, and ESLOV connectors made integrating these components straightforward.Qwiic is an I²C interface that allows users to easily interconnect modules and sensors. The Qwiic connector is polarized, so it’s virtually impossible to insert it wrong.I²C (Inter-Integrated Circuit) is a synchronous, multi-nodes serial communication protocol developed by Philips. It uses two bidirectional open-drain lines: SDA (Serial Data) and SCL (Serial Clock), pulled up with resistors. Devices on the bus are identified by unique 7-bit or 10-bit addresses, and communication is controlled by a master device that initiates data transfers.For connectivity and cloud integration, your solution can securely transmit all inspection results, including the specific type of defect and the exact timestamp, both to the <a href="https://cloud.arduino.cc/" target="_blank">Arduino Cloud</a> and locally to the production line software management. For this, you simply leveraged the <a href="https://store.arduino.cc/products/4g-module-global" target="_blank">4G GNSS Module Global</a> (included in the Proto Kit VE Kit) to provide high-speed, reliable cellular connectivity supporting 4G and fallbacks to 2G, 3G.The 4G Module is GNSS compliant and can add relevant info on GPS and other constellations.<iframe width="640" height="360" src="https://www.youtube.com/embed/HV4GWjxejvM?&amp;wmode=opaque&amp;rel=0&amp;enablejsapi=1&amp;origin=https://www.wevolver.com" frameborder="0"></iframe> <h3>Success at Precision Parts Inc.!</h3>Sarah can now see real-time updates on a dashboard pulled up directly via Arduino Cloud for Business – not just on her computer but also on her smartphone. This allows her to monitor quality trends across shifts, get statistics on defect rates, and receive instant notifications if a critical threshold of defects is reached. The data collected in the Arduino Cloud also provides invaluable insights, allowing Precision Parts Inc. to continually refine their manufacturing processes. This specific continuous data flow will help Sarah and her team to create a database to train the AI model even further, to achieve greater accuracy over time.And what if Sarah’s boss wants to have a custom report inside his preferred tool? You’re ready for that, too. Using the APIs available in Arduino Cloud, you can simply push data from the cloud to a local database, and create such report mixing real production data with finance reporting figures.What’s more, you’ve managed to develop and iterate on this entire automated quality inspection system – from setting up the camera and deploying the AI model to programming the defect detection logic and integrating with the cloud – in days instead of weeks, or perhaps even hours instead of days. What’s next? The modular design of the Portenta Mid Carrier and Portenta Mid Carrier Proto Shield make modifying or expanding the hardware simple. So Precision Parts Inc. could seamlessly transition from this functional prototype to full-scale production with minimal redesign efforts.  This story may be fictional, but the accessibility and flexibility of Arduino’s platform is a reality for thousands of companies like yours, every day.

Come with us on a journey of integration – from prototype to production – in the narrative framework of fictional company Precision Parts Inc.

A Journey into Automatic Optical Inspection

Outdoor robots cannot run blindly. They operate across unpredictable terrains, in changing weather, and under continuous workloads. To navigate these conditions, they must process visual information with precision and reliability, which begins with integrating compute and vision systems that work together seamlessly. But that integration is rarely straightforward. Engineers often face mismatched interfaces, unstable links, and firmware conflicts that delay development and complicate field testing.AVerMedia, a global leader in AI-driven edge computing solutions, and Stereolabs, a pioneer in 3D vision technology and AI-powered perception, have partnered to address this head-on. By combining a robust edge compute platform with a high-performance 3D vision system, they offer a pre-validated, deployment-ready solution. Hardware, firmware, and software are aligned out of the box, and integration becomes a starting point. This allows robotics teams to shift their focus toward application logic and field reliability, rather than debugging cables and drivers.<h2 id="isPasted">A Partnership Built to Accelerate Robotics Deployment</h2>With integration already solved at the system level, the joint platform from AVerMedia and Stereolabs is designed for the real-world challenges of outdoor robotics. Hardware and software are co-developed and validated as a system, giving robotics teams a reliable starting point without the usual integration hurdles so they can build with confidence.This partnership is also rooted in long-term support. Both companies co-develop reference materials, align on software roadmaps, and share technical documentation to support deployment. Customers benefit from consistent Board Support Package (BSP) updates and integration resources that shorten development cycles and reduce engineering overhead.Cécile Schmollgruber, CEO of Stereolabs, described the value of this partnership clearly: “This collaboration with AVerMedia demonstrates the impact of physical AI in robotics. Combining our precise vision systems with AVerMedia's reliable edge compute platforms enables robots to understand and interact with complex environments in real time.”Together, AVerMedia and Stereolabs deliver a dependable platform that shifts the engineering team’s focus from integration tasks to application design and field validation.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1754319659759-1754319659759.png" class="fr-fic fr-dib" />(Left) The D315 board from AVerMedia. Credit: AVerMedia. (Right) The Zed X One Camera from Stereolabs. Credit: Stereolabs.<h2 id="isPasted">Under the Hood: The Validated Tech Stack</h2>Deploying robotics in the real world demands a tech stack that simply works, even under challenging conditions. AVerMedia and Stereolabs have engineered a deployment-ready platform that simplifies vision integration while delivering robust performance for advanced AI workloads.At the foundation of this solution are three core products:<ul><li>AVerMedia D315 carrier board: a production-ready edge computer for NVIDIA® Jetson AGX Orin™ module, offering multi-camera GMSL2 support, built-in TPM security, 10 GbE networking, and out-of-band power control for robust field operations</li><li>Stereolabs ZED X Series cameras: An industrial-grade vision system based on multiple GMSL stereo cameras that fuse RGB, depth, and inertial data to deliver centimeter-level 3D perception in real time</li><li>Stereolabs ZED SDK: A modular software kit that runs on the D315 and exposes depth sensing, localization, mapping, and object detection APIs. Built-in AI models handle perception tasks across logistics, agriculture, and off-road applications, allowing developers to transition from prototype to deployment without rewriting core vision functions.</li></ul>Together, these three components create a plug-and-play stack that enables engineers to focus on application logic rather than low-level integration.<h2>Engineering for Reliability in the Field</h2>The engineering teams co-designed firmware, power sequencing, and BSPs to ensure every interface behaves predictably. During joint validation, they discovered an I²C address overlap between the camera and the audio codec. They resolved this by adjusting the power-on timing and then ran cold- and warm-boot cycles to confirm stability. This process produced a single software image that recognized each camera the moment it was connected, eliminating the need for manual driver work.To ensure readiness beyond the lab, the joint roadmap covers thermal validation, EMI testing, and mechanical fit checks to ensure the system performs reliably in outdoor enclosures and mission-critical environments. Both companies maintain a shared issue-tracking system, so customers receive coordinated fixes and updates. Reference designs, pinout diagrams, and ROS 2 launch files are published in a single developer portal, streamlining evaluation.<h2>Streamlining Robotics Integration</h2>The D315 platform has already been adopted in use cases such as traffic monitoring, autonomous security patrols, and smart retail analytics. Feedback from these pilots drove improvements to cable strain relief, connector sealing, and on-board voltage regulation. The production release now ships with environmental ratings that match typical robot IP enclosures. These deployments show this solution’s capability to support real-world AI applications across various environments.Paired with the ZED SDK, the platform offers a full software stack for spatial understanding. The SDK provides depth sensing, object detection, body tracking, and mapping, all powered by TERRA AI. It fuses vision, inertial, and optional GNSS data to deliver real-time 3D insights, even in harsh conditions such as low light or rain. Engineers can activate only the features they need, keeping the runtime efficient and focused on the application.For AI and robotic teams, this platform means faster starts and fewer integration headaches. Engineers can begin system development with a pre-configured stack that removes early-stage hardware and software integration work. They simply mount the camera, seat the Jetson module, flash the image, and start building autonomy logic. Time previously spent on board preparation and interface debugging now shifts to perception tuning, navigation, and application testing activities that move a robot from prototype to revenue.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1754476426918-Untitled+design+%2822%29.png" style="width:100%px" class="fr-fic fr-dib" />The AVerMedia D315 board showing its I/O interfaces. Credit: AVerMedia<h2>From Lab to Field: Star Robotics and Autonomous Security</h2>This platform has already proven itself in the field. One real-world application is Star Robotics’ Watchbot, an autonomous security and inspection robot designed for continuous operation across industrial facilities, critical infrastructure, and commercial sites. It illustrates how the D315 + ZED X + ZED SDK performs in service. Watchbot conducts round-the-clock surveillance and inspection with the mission to automate routine patrols, detect safety hazards early, and reduce human exposure in high-risk areas.Watchbot’s perception suite centers on a Stereolabs ZED X stereo camera linked to an AVerMedia D315 carrier board that hosts the robot’s Jetson compute. ROS2 orchestrates navigation and sensor fusion, while NVIDIA DeepStream streams live video for remote monitoring and AI inference.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1754320467868-1754320467868.png" class="fr-fic fr-dib" />Star Robotics’ fleet of 60+ Watchbot units has logged over 65,000 hours and 28,000 km of autonomous patrols across industrial and critical sites. Source: Star Robotics.The new setup routes power and data through a single automotive-grade GMSL2 cable, giving Watchbot a stable link between the D315 board and the ZED X camera. On that same board, the ZED SDK converts the stereo feed into depth maps, object tracks, and localisation data that the navigation stack can use immediately. Working alongside LiDAR and a 360° observation module with PTZ and thermal imaging, the Avermedia–Stereolabs combination now anchors the robot’s sensing suite.Since its first trials, Watchbot has grown into a fleet of over 60 autonomous units. Collectively, they have logged 28,000 kilometres of driver-free travel and over 65,000 hours of surveillance at factories, airports, and critical facilities. Customers rely on this system because the integrated D315, ZED X, and ZED SDK stack keeps each patrol on route and every camera stream online, day after day, even in rain, fog, and direct sunlight.<h2>Designed to Support Robotics Teams</h2>The AVerMedia and Stereolabs platform is built not only for performance but also for the pace and realities of modern robotics workflows. From evaluation to deployment, it gives engineers a stable foundation to build on without losing time to low-level system integration.Plug-and-play compatibility between the ZED X camera, the D315 carrier board, and the ZED SDK removes one of the most time-consuming steps in robotics development. Engineers no longer need to source custom drivers, patch SDKs, or modify BSPs just to establish reliable communication between camera and compute. The system boots into a working configuration with pre-validated firmware and software compatibility already in place.This integration support extends across the entire development lifecycle. Both companies maintain aligned software roadmaps, coordinated BSP release schedules, and a shared set of resources, including sample applications, reference diagrams, and ROS 2 launch files, not to mention a joint issue-tracking system for rapid fixes. Customers can also engage directly with the engineering teams to troubleshoot camera initialization, optimize firmware settings, or review power sequencing.Post-deployment, teams continue to receive updates aligned with Jetson software releases and camera firmware upgrades. This reduces the need for internal workarounds and ensures long-term maintainability in production systems. Whether building a first prototype or scaling into production, robotics teams benefit from fewer surprises, faster iteration, and more time focused on core autonomy logic. This stack scales smoothly from early prototypes to full production fleets, supporting multiple use cases like outdoor robots, mobile inspection units, and autonomous vehicles.<h2>Showcased at GTC Paris and Beyond</h2>The AVerMedia–Stereolabs collaboration was on full display at NVIDIA’s GTC Paris 2025, where the companies unveiled their integrated multi-camera AI vision platform for robotics as part of the broader Jetson ecosystem. Designed for both autonomous mobile robots (AMRs) and humanoid systems, the platform demonstrated how synchronized 3D perception and compute can be delivered in a compact, rugged form factor.It showcased real-time multi-camera streaming and depth perception running on NVIDIA Jetson, based on real-world deployments, handling AI workloads for navigation, object detection, and safety, confirming its field readiness.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1754320560388-AVerMedia-PR_06_13_2025.webp" style="width:100%px" class="fr-fic fr-dib" />AVerMedia and Stereolabs at the NVIDIA GTC 2025 event. Credit: AVerMediaThe event brought together AI developers, robotics engineers, and ecosystem partners seeking tools to bridge the gap between prototyping and deployment. The D315 + ZED X + ZED SDK platform answered that call, showing that vision integration no longer needs to be an obstacle.The <a href="https://www.avermedia.com/au/professional/media/news-detail?slug=Breakthrough-in-Vision-for-AMRs-and-Humanoid-Robots-AVerMedia-and-Vision-Pioneer-Stereolabs-Unveil-AI-Multi-Camera-Solution-at-GTC-Paris">GTC showcase</a> also positioned the solution as commercial-ready for robotics OEMs. With validated BSPs, robust system integration, and out-of-the-box compatibility, it is now part of the broader NVIDIA Jetson ecosystem roadmap. Support for next-generation Jetson modules, expanded camera configurations, and deeper ROS 2 integration is already underway, driven by feedback from early deployments.For robotics developers and manufacturers, the presence at GTC confirmed the maturity of the platform and the depth of engineering behind it. The collaboration continues to show up where it matters: in the tools and environments that robotics teams depend on to move from prototype to production.<h2>Conclusion</h2>Integrating advanced vision and compute capabilities remains one of the most resource-intensive engineering hurdles for robotics teams building field-ready systems. AVerMedia and Stereolabs remove that barrier with a pre-validated platform and coordinated support model designed to accelerate development without compromising performance.Robotics teams can prototype faster, scale confidently, and spend more time advancing autonomy instead of resolving interface issues. Whether powering a prototype or supporting a deployed fleet, the D315 + ZED X + ZED SDK stack delivers the reliability and flexibility required for real-world robotics.To learn more, please contact AVerMedia at <a>Service.AverAI@avermedia.com</a> or Stereolabs at <a>support@stereolabs.com</a>.

Star Robotics’ Watchbot, an autonomous robot with a ZED X camera and D315 board, performs 24/7 surveillance and hazard detection. Credit: Star Robotics.

Outdoor robots operate in unpredictable terrain, shifting weather, and continuous workloads. Reliable navigation requires precise visual processing enabled by seamlessly integrated compute and vision systems.

Solving Vision Integration at the Edge: Seamless AI for Robotics

Manufacturing is contending with rising material costs, increasingly complex components, and a projected shortage of 400,000 skilled workers by 2024. Manual quality control typically samples under 5% of production output, leaving room for costly defects.&nbsp;However, investments in vision AI—particularly no‑code, edge-ready platforms—offer compelling ROI, often recouped within 12–24 months. As 77% of manufacturing leaders recognize AI as a competitive advantage, no‑code platforms are gaining rapid adoption for democratizing advanced inspection capabilities without the need for data science expertise.<h2 dir="ltr">Assembly‑Sequence Verification: Zero‑Defect Builds with Matroid</h2>Matroid’s no-code computer vision platform leverages intuitive “detector studios” to accelerate deployment—no coding skills required. Notable results include:<ul><li dir="ltr">Service Center Metals: safety compliance improved from under 25% to over 90%.</li><li dir="ltr">Mercedes‑Benz is praised for enabling “rapidly deployed, highly advanced inspection solutions” that are aligned with their quality standards.<a href="https://www.matroid.com/?utm_source=chatgpt.com">&nbsp;</a></li><li dir="ltr">Steel production: defect detection accuracy reached&nbsp;<a href="https://www.matroid.com/case-studies/ai-defect-detection-steel/">98%</a> versus 60–70% manually, yielding annual savings exceeding $2M.</li><li dir="ltr">Bosch benchmarks: vision AI reduced inspection costs by 50% and achieved up to 99.8% accuracy. Vision systems detect subtle anomalies—thread engagement, alignment, torque—that human inspection often misses.</li></ul>Matroid’s camera-agnostic architecture integrates with existing hardware and processes incoming video feeds in real time to detect and notify deviations immediately.<h3 dir="ltr">Vision Analytics to Halt Cycle-Time Drift</h3>Cycle-time drift silently erodes manufacturing efficiency as processes gradually deviate from optimal performance through machine degradation and parameter changes. Vision-based timestamping provides precise production monitoring, capturing temporal data with millisecond accuracy while maintaining visual context for comprehensive analysis.&nbsp;Toyota's partnership with vision AI providers demonstrates the financial impact: preventing just one minute of downtime saves $1,000, validating how minor improvements compound into substantial returns. Manufacturing environments achieving world-class 30-40% Manufacturing Cycle Efficiency still have significant optimization potential. Vision systems enable real-time process control by detecting drift immediately rather than waiting for downstream quality checks.&nbsp;A validated methodology shows how one-second cycle reductions generate significant annual savings—a manufacturer producing 100,000 units annually with a $5 profit margin gains $50,000 from a one-second improvement across their line. Automotive implementations report 25% faster completion times for complex assemblies through vision-guided optimization.&nbsp;The technology integrates seamlessly with Manufacturing Execution Systems via OPC UA protocols, enabling centralized monitoring across multiple production lines while maintaining sub-second response times critical for real-time adjustment.<h3 dir="ltr">Dual Role: Inspect Parts and Illuminate Processes</h3>World Economic Forum research reveals augmented-workforce technologies deliver 300% ergonomics improvements and 50% quality increases by enhancing rather than replacing human capabilities. Vision AI serves this dual mandate—inspecting products while illuminating process improvements that benefit workers directly.&nbsp;The technology reduces repetitive strain by automating tedious visual inspections, freeing operators for higher-value tasks requiring human judgment and creativity. According to WEF studies, Matroid's platform includes worker-augmentation modules that provide real-time guidance through visual overlays, reducing training time by 70%.&nbsp;Standard Operating Procedure verification ensures consistent execution while capturing best practices from experienced operators. Computer vision identifies ergonomic risks, suggesting workstation adjustments that prevent injuries.&nbsp;This human-centric approach addresses manufacturing's skills gap by making expertise accessible to newer workers through AI-guided assistance. The technology transforms tribal knowledge into systematic improvements, creating feedback loops that continuously enhance product quality and worker satisfaction.<h2 dir="ltr">Implementation Blueprint</h2>Successful vision AI deployment follows a structured approach, maximizing ROI while minimizing disruption. Camera placement requires strategic positioning—high-resolution units for detailed inspection, wide-angle cameras for process monitoring.Matroid's no-code detector studio enables rapid model development: upload sample images, annotate defects or assembly states, then train custom detectors without programming.&nbsp;Edge deployment using NVIDIA Jetson or similar hardware provides millisecond response times essential for real-time control, while cloud connectivity enables model updates and enterprise-wide analytics. Integrating existing systems leverages OPC UA standards, ensuring interoperability with PLCs, SCADA, and MES platforms.The companion specification VDMA 40100 standardizes machine vision communication, simplifying deployment. Change management proves critical—the 10-20-70 rule allocates 10% effort to algorithms, 20% to infrastructure, and 70% to people and processes.&nbsp;Pilot implementations should target high-impact, well-defined use cases demonstrating clear ROI within 90 days. Success requires cross-functional teams including operations, quality, IT, and frontline workers who understand pain points intimately.<h3 dir="ltr">Business Case &amp; KPIs</h3>Vision AI delivers measurable returns across multiple vectors. Bosch's implementations validate 50% quality cost reductions through automated inspection and 20-30% energy savings via optimized processes.Typical manufacturing deployments achieve 100-350% ROI over three years with 12-24 month payback periods. Key performance indicators include: defect detection rates (target 95 %+), false favorable rates (under 2%), cycle time improvements (10-30%), and inspection cost per unit (50% reduction).Additional benefits encompass reduced scrap, lower rework costs, decreased warranty claims, and improved customer satisfaction scores. Investment requirements range from $200-500K, including cameras, edge computing, and integration, with $20-60K annual maintenance costs. The business case strengthens when considering labor shortage impacts—vision systems operate continuously without fatigue, providing consistent quality assurance as workforce availability tightens.<h2 dir="ltr">Looking Ahead</h2>Edge-native AI represents the future of manufacturing intelligence, with markets growing from $20.78 billion to $66.47 billion by 2030. Digital twin integration creates closed-loop systems where vision data feeds virtual models, enabling predictive optimization. Multi-agent AI systems will coordinate complex manufacturing processes, with 50% of enterprises deploying AI agents by 2027.&nbsp;Sustainability benefits multiply as vision systems reduce material waste and energy consumption—AI-driven optimization cuts CO2 emissions by 12-44%, depending on application. Small and medium enterprises gain unprecedented access to advanced capabilities through cloud-based platforms and government support initiatives.&nbsp;By 2026, ecosystem developments will democratize vision AI, making enterprise-grade inspection accessible to manufacturers of all sizes. The convergence of 5G, edge computing, and generative AI will enable real-time adaptation to changing conditions, moving beyond fixed inspection routines to dynamic quality optimization.<h2 dir="ltr">Conclusion</h2>Vision AI transforms manufacturing quality control from reactive sampling to proactive, comprehensive inspection. No-code platforms like Matroid eliminate technical barriers, enabling rapid deployment of sophisticated inspection systems that serve as intelligent co-pilots for human workers.&nbsp;The technology simultaneously improves product quality, worker ergonomics, and operational efficiency while delivering compelling ROI. Ready to explore how vision AI can eliminate assembly errors and cycle drift in your operations?&nbsp;Schedule a discovery call to see Matroid's&nbsp;<a href="https://www.matroid.com/">platform</a> and discuss your manufacturing challenges. Our team will demonstrate relevant use cases and help design a pilot program targeting your highest-impact quality issues.<h3 dir="ltr">References</h3><ol><li dir="ltr">Thomasnet&nbsp;- Supplier Defect Rate Insights <a href="https://www.thomasnet.com/insights/supplier-defect-rate/">https://www.thomasnet.com/insights/supplier-defect-rate/</a></li><li dir="ltr">National Association of Manufacturers&nbsp;- Supply Chain and Workforce Concerns<a href="https://nam.org/supply-chain-rising-raw-material-costs-and-workforce-shortages-top-concerns-for-manufacturers-17080/">&nbsp;https://nam.org/supply-chain-rising-raw-material-costs-and-workforce-shortages-top-concerns-for-manufacturers-17080/</a></li><li dir="ltr">Manufacturing Tomorrow&nbsp;- Labor Shortages in Metal Fabrication <a href="https://www.manufacturingtomorrow.com/story/2022/02/how-will-labor-shortages-impact-metal-fabrication-in-2022/18341/">https://www.manufacturingtomorrow.com/story/2022/02/how-will-labor-shortages-impact-metal-fabrication-in-2022/18341/</a></li><li dir="ltr">InTouch Quality&nbsp;- Quality Control Inspection Levels<a href="https://www.intouch-quality.com/blog/quality-control-inspection-3-general-inspection-levels">&nbsp;https://www.intouch-quality.com/blog/quality-control-inspection-3-general-inspection-levels</a></li><li dir="ltr">Insight Quality&nbsp;- AQL Sampling Guidelines <a href="https://insight-quality.com/what-is-aql-and-what-do-you-need-to-know-about-it/">https://insight-quality.com/what-is-aql-and-what-do-you-need-to-know-about-it</a></li><li dir="ltr">ClearObject&nbsp;- Vision AI for Quality Control <a href="https://www.clearobject.com/vision-ai-for-quality-control/">https://www.clearobject.com/vision-ai-for-quality-control/</a></li><li dir="ltr">McKinsey &amp; Company&nbsp;- The State of AI <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai">https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai</a></li><li dir="ltr">Dialpad&nbsp;- State of AI and C-Suite Report <a href="https://www.dialpad.com/blog/ai-csuite-report/">https://www.dialpad.com/blog/ai-csuite-report/</a></li><li dir="ltr">Matroid&nbsp;- No-Code Computer Vision Platform <a href="https://www.matroid.com/">&nbsp;https://www.matroid.com/</a></li><li dir="ltr">Matroid&nbsp;- Computer Vision in Manufacturing <a href="https://www.matroid.com/computer-vision-manufacturing/">https://www.matroid.com/computer-vision-manufacturing/</a></li><li dir="ltr">Matroid&nbsp;- Advanced Computer Vision Solutions<a href="https://www.matroid.com/advanced-computer-vision/">&nbsp;https://www.matroid.com/advanced-computer-vision/</a></li><li dir="ltr">Matroid&nbsp;- Key Applications in Industrial Manufacturing<a href="https://www.matroid.com/key-applications-of-computer-vision-in-industrial-manufacturing/">&nbsp;https://www.matroid.com/key-applications-of-computer-vision-in-industrial-manufacturing/</a></li><li dir="ltr">Matroid&nbsp;- Why Choose Matroid <a href="https://www.matroid.com/why-matroid/">&nbsp;https://www.matroid.com/why-matroid/</a></li><li dir="ltr">Matroid&nbsp;- Workplace Safety Monitoring Case Study <a href="https://www.matroid.com/case-studies/how-matroids-ai-computer-vision-can-transform-workplace-safety/">https://www.matroid.com/case-studies/how-matroids-ai-computer-vision-can-transform-workplace-safety/</a></li><li dir="ltr">Matroid&nbsp;- Steel Slab Defect Detection Case Study <a href="https://www.matroid.com/case-studies/ai-defect-detection-steel/">https://www.matroid.com/case-studies/ai-defect-detection-steel/</a></li><li dir="ltr">Wevolver&nbsp;- Computer Vision for Safety and Quality Challenges <a href="https://www.wevolver.com/article/leveraging-computer-vision-to-tackle-safety-and-quality-challenges-in-manufacturing">https://www.wevolver.com/article/leveraging-computer-vision-to-tackle-safety-and-quality-challenges-in-manufacturing</a></li><li dir="ltr">Taylor &amp; Francis&nbsp;- Manufacturing Cycle Time Research<a href="https://www.tandfonline.com/doi/abs/10.1080/00207540701513893">&nbsp;https://www.tandfonline.com/doi/abs/10.1080/00207540701513893</a></li><li dir="ltr">TWI Institute&nbsp;- Reducing Cycle Time <a href="https://www.twi-institute.com/reduce-cycle-time/">https://www.twi-institute.com/reduce-cycle-time/</a></li><li dir="ltr">Katana&nbsp;- Manufacturing Cycle Time Analysis <a href="https://katanamrp.com/blog/manufacturing-cycle-time/">https://katanamrp.com/blog/manufacturing-cycle-time/</a></li><li dir="ltr">Symestic&nbsp;- Manufacturing Cycle Efficiency <a href="https://www.symestic.com/en-us/what-is/manufacturing-cycle-efficiency">https://www.symestic.com/en-us/what-is/manufacturing-cycle-efficiency</a></li><li dir="ltr">Appinventiv&nbsp;- Computer Vision in Manufacturing Use Cases <a href="https://appinventiv.com/blog/computer-vision-in-manufacturing/">https://appinventiv.com/blog/computer-vision-in-manufacturing/</a></li><li dir="ltr">Harvard Business Review&nbsp;- Augmented Reality Improving Worker Performance <a href="https://hbr.org/2017/03/augmented-reality-is-already-improving-worker-performance">https://hbr.org/2017/03/augmented-reality-is-already-improving-worker-performance</a></li><li dir="ltr">IBM&nbsp;- AI in Manufacturing<a href="https://www.ibm.com/think/topics/ai-in-manufacturing">&nbsp;https://www.ibm.com/think/topics/ai-in-manufacturing</a></li><li dir="ltr">OPC7&nbsp;- OPC UA and MES Integration <a href="https://www.opc7.com/opc-ua-and-mes/">&nbsp;https://www.opc7.com/opc-ua-and-mes/</a></li><li dir="ltr">ResearchGate&nbsp;- World Economic Forum Augmented Workforce Paper <a href="https://www.researchgate.net/publication/358037363_Augmented_Workforce_Empowering_People_Transforming_Manufacturing_World_Economic_Forum_White_Paper">https://www.researchgate.net/publication/358037363_Augmented_Workforce_Empowering_People_Transforming_Manufacturing_World_Economic_Forum_White_Paper</a></li><li dir="ltr">Workable&nbsp;- Augmented Workforce Implementation <a href="https://resources.workable.com/stories-and-insights/augmented-workforce-is-happening-now">https://resources.workable.com/stories-and-insights/augmented-workforce-is-happening-now</a></li><li dir="ltr">KIT-AR&nbsp;- Augmented Workers in Manufacturing<a href="https://kit-ar.com/article-augmented-workers-in-manufacturing/">&nbsp;https://kit-ar.com/article-augmented-workers-in-manufacturing/</a></li><li dir="ltr">LightGuide Systems&nbsp;- Digitally Connected Worker <a href="https://www.lightguidesys.com/resource-center/blog/the-digitally-connected-worker-your-manufacturing-workforce-of-the-future/">https://www.lightguidesys.com/resource-center/blog/the-digitally-connected-worker-your-manufacturing-workforce-of-the-future/</a></li><li dir="ltr">Autodesk&nbsp;- Augmented Reality in Manufacturing <a href="https://www.autodesk.com/solutions/augmented-reality">https://www.autodesk.com/solutions/augmented-reality</a></li><li dir="ltr">Scale Computing&nbsp;- Edge AI GPU Performance <a href="https://www.scalecomputing.com/resources/how-gpu-are-transforming-edge-ai-for-real-time-performance">https://www.scalecomputing.com/resources/how-gpu-are-transforming-edge-ai-for-real-time-performance</a></li></ol>

Matroid's no-code computer vision platform leverages an intuitive 'detector studio' to accelerate deployment - no coding skills required.

No-Code Vision AI: Ending Assembly Errors and Cycle Drift

Detecting defects on semiconductor wafers is extremely challenging due to their reflective surfaces, micro-scale defects, and the need for high precision. Manual parameter tuning and traditional rule-based approaches can be time-consuming and inconsistent, especially when scaling across varying product lines. Machine vision software is therefore playing an increasingly critical role in semiconductor manufacturing, by enabling automated, robust, and highly efficient optical inspection processes.Join MVTec on June 5th for a step-by-step, demo-driven webinar and discover how advanced vision technologies are transforming wafer inspection and driving quality as well as throughput in semiconductor manufacturing. <a href="https://www.mvtec.com/application-areas/semiconductors/on-demand-webinar-how-to-maximize-the-yield-in-wafer-production" target="_blank" rel="noopener noreferrer" class="button-main-action">VIEW THE ON-DEMAND WEBINAR</a><h2> </h2>Discover a smarter way to detect wafer defects – no slides, just solutions.<h3>What You’ll Learn</h3><ul><li>How Machine Vision can be used to&nbsp;detect defects on a wafer&nbsp;showing a&nbsp;live demo</li><li>Which&nbsp;technologies&nbsp;in&nbsp;MVTec HALCON&nbsp;can be used to&nbsp;maximize the yield</li><li>How MVTec can support you to start with an automated optical inspection</li></ul><h3> </h3> <table><tbody><tr><td><img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzQ2NzAzOTIyMjE3LW1zZV8yLmpwZWciLCJlZGl0cyI6eyJyZXNpemUiOnsid2lkdGgiOjk1MCwiZml0IjoiY292ZXIifX19" style="width:486px;" class="fr-fic fr-dib"></td><td>Moderator<a href="https://www.linkedin.com/in/markus-setzer-9113731b7/">Markus Setzer&nbsp;</a>has been a Marketing Specialist at MVTec for six years. In his role at MVTec, he talks a lot with customers and interested parties. Through trade fairs in particular, he has a deep insight into the industry's most pressing questions regarding machine vision in production. The demands of the market shape the structure of the webinar, which Markus Setzer will moderate.</td></tr><tr><td>Application Engineer<a href="https://www.linkedin.com/search/results/all/?fetchDeterministicClustersOnly=true&heroEntityKey=urn%3Ali%3Afsd_profile%3AACoAADfmAmIB6H7cwuWEOcdS5WcNGGUXKD81pmI&keywords=coco%20rogers&origin=RICH_QUERY_TYPEAHEAD_HISTORY&position=0&searchId=4e8844cd-9d7e-4dc1-b3b1-8a2f60079e25&sid=YTi&spellCorrectionEnabled=true">Coco Rogers</a> works as an Application Engineer in the Customer Service department and brings extensive expertise in defect detection. She has in-depth knowledge of the semiconductor industry and also develops demonstrators that showcase our software in action. During the webinar, Coco will give a live demonstration of how the optical inspection process of a wafer is performed using advanced machine vision technologies.</td><td><img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzQ2NzAzODg1MTAyLUNvY29fV2ViaW5hci5qcGciLCJlZGl0cyI6eyJyZXNpemUiOnsid2lkdGgiOjk1MCwiZml0IjoiY292ZXIifX19" style="width:100%px;" class="fr-fic fr-dib"></td></tr></tbody></table><h2> </h2><a href="https://www.mvtec.com/application-areas/semiconductors/on-demand-webinar-how-to-maximize-the-yield-in-wafer-production" target="_blank" rel="noopener noreferrer" id="isPasted" class="button-main-action">VIEW THE ON-DEMAND WEBINAR</a><h2 id="isPasted">About MVTec</h2>MVTec develops machine vision software used across a wide range of industrial applications, including semiconductor manufacturing. Its tools support tasks such as wafer alignment, inspection of dies and ball/wire bonds, defect classification, shape matching, precise measurement in 2D and 3D, as well as 3D robot guidance and identification. These capabilities contribute to improving production efficiency, enabling advanced manufacturing processes, and supporting applications where human-machine collaboration is required.

Join MVTec to explore how advanced machine vision software is revolutionizing semiconductor wafer inspection with live demos, not slides—boosting precision, quality, and efficiency.

How to maximize the yield in wafer production webinar

<h2 id="isPasted">Introduction</h2>Aluminum extrusion is a vital manufacturing process that shapes aluminum into components used in aerospace, automotive, and construction industries. Driven by its lightweight, strength, and recyclability, the global market for aluminum extrusion was valued at $93.4 billion in 2023 and is projected to reach $198.1 billion by 2033, reflecting its growing importance in sustainable and innovative application. However, despite its efficiency, the process faces challenges such as speed tears, which are small fractures that can compromise product integrity and lead to increased costs and material waste.In today's competitive manufacturing landscape, traditional manual inspection methods for aluminum extrusion are increasingly inadequate due to their time-consuming nature, susceptibility to human error, and inability to detect defects like speed tears in real-time. As the demand for high-quality aluminum products grows, manufacturers increasingly turn to advanced technologies like artificial intelligence (AI) and computer vision to streamline operations.Matroid, a leader in AI-powered computer vision solutions, has emerged as a leader in this field. Their automated speed tear detection system offers a cutting-edge solution to one of the most pressing challenges in aluminum extrusion. This article explores how Matroid’s technology transforms quality control in metal manufacturing and why it represents the industry's future.<h2>The Challenges of Quality Control in Aluminum Extrusion</h2>The aluminum extrusion process is complex, with quality influenced by factors such as alloy composition, temperature regulation, extrusion speed, and die design. Maintaining uniform quality is challenging due to defects like surface scratches, die lines, and, most critically, speed tears. These tears are caused by excessive friction or localized stress during extrusion. They are particularly problematic as they can develop unexpectedly and spread rapidly, leading to significant material loss if not identified promptly. The fast pace of production lines and the subtle nature of these defects further complicate detection efforts.Historically, manual inspection has been the primary method for identifying defects. However, this approach is labour-intensive, prone to human error, and inconsistent due to operator fatigue and subjective judgment. Traditional methods also struggle with false positives and lack systematic data collection to correlate defects with process parameters, making it difficult to identify root causes and implement preventative measures. These limitations increase production costs and highlight the need for more robust quality control systems to ensure high-quality aluminum extrusion products while minimizing waste and inefficiencies.<h2>Matroid’s AI-Powered Solution: Redefining Defect Detection for Modern Manufacturing</h2>Matroid, a leader in computer vision solutions, collaborated with a leading aluminum extruder and offered an innovative approach to speed tear detection using artificial intelligence. The AI-powered speed tear detection system is a prime example of how computer vision can address the challenges of aluminum extrusion. The system analyses real-time video footage of the extrusion process, identifying speed tears as they occur. This real-time capability is critical for minimizing waste and identifying defects before they escalate into more significant issues.Optimized lighting techniques were implemented to enhance the system’s effectiveness and improve defect contrast, making it easier for AI to detect even subtle imperfections. Additionally, Matroid collaborated closely with the aluminum extruder to capture operator expertise. Sourcing and defining various speed and hot tear profiles created a plant-wide standard for all packaging operators, metallurgists, and process engineers. This collaboration ensured the AI system was trained on accurate and representative data.One of the standout features of Matroid’s solution is its adaptive AI detector. The system was trained on a diverse dataset encompassing various extrusion profiles, ensuring it could adapt to different shapes and configurations. This adaptability is crucial for manufacturers dealing with various product types. Furthermore, Matroid employed a smart annotation strategy, distinguishing between actual defects and non-defect patterns using newly created “tear definitions.” This approach allowed the detector to quickly differentiate critical tears with minimal training data, significantly speeding up the implementation process.Matroid’s solution's accuracy is one of its key advantages. The system can detect even the most minor defects, ensuring that only high-quality products move forward in the production line. Additionally, the system is highly scalable, making it suitable for manufacturers of all sizes. Matroid’s system is also designed with ease of integration in mind. The solution can be seamlessly incorporated into existing production lines, minimizing disruption and maximizing Region of Interest (ROI). This makes it an attractive option for manufacturers looking to modernize their operations without overhauling their infrastructure. By utilizing Matroid's computer vision technology, manufacturers can improve product quality, achieve substantial cost savings, and enhance operational efficiencies.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1746626166131-1746626166131.png" class="fr-fic fr-dib" /><h2>Conclusion</h2>Matroid’s AI-powered computer vision solution marks a significant advancement in quality control for aluminum extrusion manufacturing. The system’s ability to detect speed tears in real time, combined with optimized lighting techniques and adaptive AI algorithms, addresses long-standing challenges in the industry. Its seamless integration into existing production lines, scalability, and user-friendly design make it a powerful tool for manufacturers seeking to reduce waste, cut costs, and enhance product quality.AI and computer vision are poised to play a transformative role in manufacturing in the coming years. As Industry 4.0 evolves, integrating predictive analytics and Internet of Things (IoT) technologies will significantly enhance defect detection and prevention. These advancements will enable manufacturers to transition from reactive quality control to proactive strategies, effectively minimizing defects before they occur.Additionally, as global emphasis on sustainability increases, technologies like Matroid’s will become indispensable in reducing material waste and optimizing resource efficiency. Leveraging AI-driven solutions allows manufacturers to improve operational efficiency while contributing to a more sustainable and innovative manufacturing ecosystem. Matroid’s pioneering approach is solving current challenges and setting the stage for a smarter, more efficient future in aluminum extrusion.To learn more about how Matroid’s AI-driven solutions can enhance your manufacturing operations, visit <a href="https://www.matroid.com/">Matroid’s website</a>.<hr /><h1 id="isPasted">References:</h1><ol><li><a href="https://simmal.com/the-advantages-of-aluminium-extrusion-in-manufacturing/">https://simmal.com/the-advantages-of-aluminium-extrusion-in-manufacturing/</a></li><li><a href="https://vocal.media/01/aluminum-extrusion-a-vital-manufacturing-process-driving-innovations">https://vocal.media/01/aluminum-extrusion-a-vital-manufacturing-process-driving-innovations</a></li><li><a href="https://www.verifiedmarketresearch.com/product/aluminum-extrusion-market/">https://www.verifiedmarketresearch.com/product/aluminum-extrusion-market/</a></li><li><a href="https://www.alcircle.com/news/tech-trends-driving-the-aluminium-extrusion-industry-111131">https://www.alcircle.com/news/tech-trends-driving-the-aluminium-extrusion-industry-111131</a></li><li><a href="https://www.alliedmarketresearch.com/aluminum-extrusion-market-A09001#:~:text=What%20is%20the%20estimated%20industry,7.8%25%20from%202024%20to%202033">https://www.alliedmarketresearch.com/aluminum-extrusion-market-A09001#:~:text=What%20is%20the%20estimated%20industry,7.8%25%20from%202024%20to%202033</a>.</li><li><a href="https://www.eitmanufacturing.eu/news-events/news/ai-driven-in-line-quality-assurance-for-aluminium-extrusion/">https://www.eitmanufacturing.eu/news-events/news/ai-driven-in-line-quality-assurance-for-aluminium-extrusion/</a></li><li><a href="https://aimlprogramming.com/services/ai-enabled-aluminum-extrusion-monitoring/">https://aimlprogramming.com/services/ai-enabled-aluminum-extrusion-monitoring/</a></li><li><a href="https://www.matroid.com/leveraging-computer-vision-to-improve-efficiency-quality-control-and-workplace-safety-in-metal-forming/">https://www.matroid.com/leveraging-computer-vision-to-improve-efficiency-quality-control-and-workplace-safety-in-metal-forming/</a></li><li><a href="https://www.matroid.com/computer-vision-in-metal-forming-precision-efficiency-and-roi/">https://www.matroid.com/computer-vision-in-metal-forming-precision-efficiency-and-roi/</a></li><li><a href="https://www.matroid.com/case-studies/how-matroids-ai-computer-vision-can-transform-workplace-safety/">https://www.matroid.com/case-studies/how-matroids-ai-computer-vision-can-transform-workplace-safety/</a></li><li><a href="https://aluminumextrusions.net/aluminum-extrusion-quality-control/">https://aluminumextrusions.net/aluminum-extrusion-quality-control/</a></li><li><a href="https://globalaluminium.com/quality-of-aluminium-extrusion-manufacturing/">https://globalaluminium.com/quality-of-aluminium-extrusion-manufacturing/</a></li><li><a href="https://www.chaluminium.com/common-aluminum-extrusion-defects-causes-and-solutions">https://www.chaluminium.com/common-aluminum-extrusion-defects-causes-and-solutions</a></li><li><a href="https://aluphant.com/quality-control-measures-in-aluminum-extrusion-manufacturing-a-comprehensive-overview/">https://aluphant.com/quality-control-measures-in-aluminum-extrusion-manufacturing-a-comprehensive-overview/</a></li></ol>

Machine Vision Software

ORGANIZATIONS.

SHAPING THE INDUSTRY.

MVTec Software

Stereolabs

Emberion

Latest Posts