PROPHESEE

What a system ignores can be as important as what it sees. In defense and aerospace environments, perception systems are often asked to do more with less: less bandwidth, less power, less room for additional payload, less time, and less tolerance for missed information. This is especially true for drones and small moving objects, where the challenge is not only to see more, but to isolate the right signal from complex, cluttered, and visually noisy scenes. Event-based vision is built around that principle: effective perception does not always start with capturing everything. When each pixel responds independently to change, the system produces a sparse, motion-rich stream of information that reduces redundancy at the source, enabling substantially lower data load, reduced latency, and more power-efficient processing. With Prophesee Metavision® sensing and AI, this means ultra-high-speed perception, low-latency response, >140 dB dynamic range, and low-SWaP integration for applications where conventional visual pipelines often reach their limits – from drone detection and tracking to GPS-denied navigation, obstacle avoidance and more. The future of defense and aerospace perception will not be built by capturing everything. It will be built by capturing what matters. See how Prophesee is enabling real-time perception for defense and aerospace systems: https://lnkd.in/d8RDAhaU #DefenseTech #Aerospace #CounterUAS #EdgeAI #MachineVision #EventBasedVision

Most structural issues don’t announce themselves. They build up in the background – through small, fast changes that rarely get captured when systems are checked intermittently. Seeing that evolution requires a different kind of signal. Event-based vision systems capture changes as they happen and turn them into a continuous stream of information about motion and behavior – preserving micro-motion, handling ultra-high-speed dynamics, and maintaining robustness in extreme lighting or high-contrast environments where detail is often lost. Interpreted over time, these signals provide a more consistent understanding of how a system behaves – not just at inspection points, but between them – helping isolate meaningful patterns and support earlier, more informed decisions. 🛤️ In railway infrastructure, this can directly impact operations, reducing disruption, limiting emergency interventions, and preventing losses that can reach millions within a single network. Catenary systems operate under constant mechanical stress, where early signs of instability or wear can develop between inspection cycles. Tracking these changes as they emerge opens up a more continuous, real-time view of system behavior under real conditions. ➡️ See how this can be evaluated in practice: https://lnkd.in/d9wxUnPN #MachineVision #IndustrialAI #EdgeAI #PredictiveMaintenance #StructuralMonitoring

Robotics is starting to operate in timeframes that were previously out of reach. What’s changing is not just the learning approach, but how quickly a system can perceive and react to the physical world. In high-speed interaction tasks like table tennis, even small delays or inaccuracies in sensing lead to missed or unstable actions, as trajectories evolve rapidly and subtle variations – like spin – determine the outcome of each exchange. Sony's project Ace pushes this to the limit. Tracking and returning the ball requires perception that captures motion with both precision and continuity, fast enough to stay aligned with the dynamics of the game. Event-based vision is part of how this is achieved, capturing changes as they happen and preserving the timing needed to follow fast trajectories and infer spin without delay. In this system, the IMX636 event-based sensors – developed in collaboration between Sony and Prophesee – contribute to enabling this level of responsiveness. The result is a perception pipeline operating at ~10 ms latency, supporting stable, real-time interaction. In systems like this, the challenge is not only understanding the scene – but doing so at the speed at which the physical world evolves. Learn more about event-based vision: https://lnkd.in/dpRxUP8E #EventBasedVision #RobotPerception #MachineVision #PhysicalAI

Between vision sensing and application, signals need to be turned into usable information in real time. With event-based vision, this is built around changes rather than processing full frames – operating on sparse, asynchronous events that preserve temporal detail while keeping data efficient. This is where raw event streams are filtered, grouped, and translated into outputs a system can act on – whether that’s tracking motion, isolating patterns, or responding as changes emerge. Environments like Metavision® SDK are part of that transformation, providing the building blocks used to develop applications for real-world systems. ⬇️ Explore what can be built with event-based vision: https://lnkd.in/dmSf4qdq #EventBasedVision #MachineVision #EdgeAI #ComputerVision #PhysicalAI

Some of the most important signals in physical systems are not directly visible – they exist as subtle, high-frequency dynamics that conventional frame-based cameras tend to average out or miss entirely. This is particularly evident with vibrations, where small, fast changes can reveal how a system behaves, degrades, or fails, long before any visible indication appears. Event-based vision is inherently suited to capturing these dynamics. By recording changes in brightness continuously and asynchronously, it enables much finer temporal resolution than frame-based approaches, supporting non-intrusive measurement even under standard lighting conditions and without complex setups. In a recent work from the University of Tsukuba, researchers show how vibration signals can be reconstructed directly from event data, demonstrating accurate recovery of both amplitude and frequency, as well as the ability to separate concurrent vibration sources. More broadly, it reflects a shift in machine perception from capturing static images to sensing the dynamics of the world in real time. This is increasingly relevant in real-world systems, including structural monitoring, industrial inspection, and early fault detection. Explore applications of event-based vision in industrial environments: https://lnkd.in/et_4nnaS Full paper: https://lnkd.in/e55ipxVg https://lnkd.in/eDicVkka   #EventBasedVision #MachineVision #EdgeAI #Perception #IndustrialAI

As industrial systems move toward higher levels of automation, vision is becoming part of the control loop – where latency and reliability directly impact system performance. In a recent article published by EDN, Thibaut Willeman, Head of BD & Go-to-Market at Prophesee, explores what this shift means for machine vision architectures. Read the full article: https://lnkd.in/dWrWmd3a EDN: Voice of the Engineer #MachineVision #IndustrialAutomation #EdgeAI

The biggest bottleneck in Physical AI isn’t AI. It’s perception 👁️ Most AI progress so far has happened in digital environments – language, images, data. But Physical AI operates in the real world: robots, autonomous machines, industrial systems, drones. In these environments, intelligence is limited by how fast and reliably these machines can perceive and react. Most machine vision still relies on frame-based cameras – technology originally designed for human viewing. Capturing full images at fixed intervals creates redundant data, latency, and processing bottlenecks for real-time applications. Event-based vision is built on a different sensing principle 💡 With event-based sensing, systems can detect changes in a scene asynchronously and continuously. The sensor outputs only what changes in a scene, when it changes. This enables ultra-low latency, high dynamic range, significantly lower data rates, and sensing systems that are inherently privacy-preserving since no images are recorded. Physical AI is also not a single technology but an ecosystem spanning robotics, sensors, semiconductors, AI software, and industrial systems. A recent Robot Report article discussing the Physical AI landscape in Europe highlights how sensing and robotics companies are becoming an important part of this emerging stack: https://lnkd.in/dWn22ZPs As these systems develop, the performance of intelligent machines will increasingly depend on how they sense and understand the world in real time. If you’re exploring real-time perception for Physical AI systems, learn more about event-based vision here: https://lnkd.in/dtqYRExH #PhysicalAI #EventBasedVision #MachineVision #EdgeAI #Robotics #AutonomousSystems #IndustrialAutomation

⏱️ In mission-critical defense and aerospace operations, every millisecond counts. Many systems must detect, track and navigate in conditions with fast motion, extreme lighting contrast, limited bandwidth or GPS-denied environments, where perception latency and reliability become system-level constraints. Event-based vision from Prophesee addresses these challenges by detecting changes in a scene continuously and asynchronously, enabling microsecond-level latency, temporal resolution equivalent to up to ~10,000 fps, ultra-high dynamic range (>140 dB), and significantly lower data rates than conventional frame-based approaches. This sensing architecture enables systems to detect and react much faster, operate reliably in challenging lighting or highly dynamic scenarios, process far less data at the edge, support efficient edge AI processing, and integrate into platforms with tight SWaP constraints. Swipe through the carousel below to explore several examples of systems and programs using Prophesee’s event-based vision - from drone detection and navigation to space situational awareness. 👉 Learn more about event-based vision for defense & aerospace: https://lnkd.in/djtnCRAf #EventBasedVision #MachineVision #DroneDetection #UAV #Defense #Aerospace

Engineering becomes more interesting when the system has to work in the real world, not just on a screen. When machines have to perceive and react in real time – in a factory, on a robot, or in an aircraft – the constraints are different. Performance is not only measured in accuracy, but also in latency, power efficiency and reliability. Hardware, embedded software and AI have to work together as one system. 🌍 At Prophesee, engineers work on technologies that don’t stay in simulations or demos, but become part of systems operating in the real environments. As a pioneer in event-based vision, Prophesee has seen its technology evolve from early research and breakthrough applications in space and medical research, including partial vision restoration, to today’s industrial, robotics and defense systems, where real-time perception is critical. Working on technology that progresses from early concepts to real-world deployment is a defining part of the engineering work at Prophesee. As these systems move into larger programs and new applications, we are growing our engineering teams in Paris and looking for engineers who want to help shape what comes next in machine vision. Explore open roles: https://lnkd.in/dZ2YbgtG #DeepTech #HardwareEngineering #MachineVision #ComputerVision #HiringEngineers

Fast drone motion quickly challenges traditional 3D reconstruction. In this work from Davide Scaramuzza’s team at the University of Zurich, event-based vision is combined with motion-blurred frames to recover sharp scene detail even under high-speed conditions. With pixels responding asynchronously to changes in light, event data captures motion at much finer temporal resolution than frame-based approaches, preserving information that would otherwise be lost to motion blur. The result of four years of work – and a meaningful step toward more reliable perception in high-speed scenarios. Great to see Prophesee’s event-based vision technology playing a role in work like this. #AutonomousSystems #EventBasedVision #3DReconstruction #DroneNavigation