Introducing Balor Defense
Our event-based vision and edge-autonomy work is now a dedicated product line focused on seeker and guidance solutions for counter-UAS interceptors, small arms and point defense.
News
Announcements & field notes
Program milestones, technical notes and company news.
Field note: what breaks in the last 500 meters
Four failure modes that compound at precisely the moment a counter-UAS engagement is resolved — and why they are sensor architecture problems, not algorithm problems.
How ICCV 2025 researchers put event cameras on the drone detection problem
A new ICCV 2025 workshop paper tackles UAS detection with event cameras directly, demonstrating that the sensor's microsecond response time gives it structural advantages over frame-based approaches for detecting fast, small aerial targets.
FPGA-accelerated neuromorphic vision for orbital object detection
Researchers demonstrate a real-time neuromorphic vision pipeline on FPGA hardware for tracking orbital objects — an architecture that transfers directly to the embedded seeker problem.
A new survey maps where event-based vision is winning in space applications
A comprehensive 2026 survey of event-based vision in space identifies the same sensor properties that drive Balor's terminal-phase work — high dynamic range, sub-millisecond latency, and noise immunity in high-radiation environments.
New SLAM benchmark exposes where event cameras outrun their algorithms
A 2026 benchmark for event-based SLAM under high-speed maneuver conditions reveals that sensor hardware has pulled ahead of estimation algorithms — relevant context for anyone building a guidance system around event data.
EdgeSpike brings spiking neural networks to low-power edge sensing
A 2026 paper demonstrates spiking neural networks running inference on constrained edge hardware at power levels below 1W — bridging the gap between neuromorphic sensing and deployable embedded autonomy.
A framework for energy-efficient neuromorphic computing on edge AI targets
A 2026 paper proposes a hardware-aware optimization framework for deploying adaptive spiking neural networks on power-constrained edge processors — with direct implications for embedded seeker design.
Deep visual odometry for stereo event cameras reaches deployment-level accuracy
Researchers demonstrate learned visual odometry from stereo event cameras that achieves positioning accuracy competitive with state-of-the-art frame-based methods — at update rates an order of magnitude higher.
ICML 2025 paper combines spiking networks with vision transformers for event-based detection
A hybrid architecture that embeds spiking layers inside a vision transformer demonstrates strong detection accuracy on event-camera data while maintaining energy efficiency — an important step toward deployable neuromorphic detection.
Stereo event cameras unlock continuous-time 3D object detection
A 2025 preprint demonstrates that stereo event cameras enable continuous-time 3D bounding-box detection — eliminating the frame-rate constraint that limits conventional 3D detection to a fixed temporal grid.
ICCV 2025 introduces the first event-camera benchmark for tiny object detection
A new ICCV 2025 benchmark specifically targets tiny object detection with event cameras — filling a critical gap in the literature for small, distant aerial target scenarios.
Mamba state-space models tackle event-guided video deblurring at ICCV 2025
EVDM introduces Mamba-based state-space models for event-guided video deblurring, achieving cleaner frame reconstruction from blurred video with fewer parameters and faster inference than transformer-based approaches.
MTevent: a multi-task event-camera dataset covering 6D pose and moving object detection
A new 2025 dataset provides synchronized event-camera data with ground truth for both 6D pose estimation and moving object detection — enabling joint training that better reflects real sensing requirements.
Propeller signatures enable reliable drone tracking from a moving event camera
ICONS 2025 paper demonstrates that propeller-induced event signatures provide a tracking feature stable enough to maintain lock on a drone from a moving camera platform — directly relevant to seeker applications.
NTIRE 2025 deblurring challenge reveals the state of event-guided image restoration
The NTIRE 2025 challenge on event-based image deblurring assembled the community around a standardized benchmark, revealing both the current ceiling for hybrid event-frame systems and the gaps that remain.
A 2025 review maps the full stack of neuromorphic vision sensor technology
A comprehensive review in Sensors covers the hardware, algorithms, and applications of neuromorphic vision sensors from pixel circuit to application — a useful reference for anyone entering the field.
On-chip learning makes neuromorphic processors adaptable in deployment
A 2025 paper demonstrates spiking neural networks with on-chip learning running on commodity neuromorphic hardware — enabling deployed sensors to adapt to new conditions without offboard training infrastructure.
BiasBench: finally a reproducible protocol for event camera bias tuning
A 2025 paper introduces the first reproducible benchmark for tuning the bias parameters of event cameras — addressing a practical barrier to deploying event cameras in consistent, controlled configurations.
Asynchronous collaborative graph representation brings frames and events together at CVPR 2025
CVPR 2025's ACGR paper proposes a unified graph representation that processes frame and event data asynchronously — enabling true temporal fusion rather than the forced synchronization that limits current hybrid approaches.
Temporal-guided spiking networks advance event-based action recognition
A 2025 paper introduces temporal guidance mechanisms for spiking neural networks that improve event-based action recognition — with implications for target behavior classification in C-UAS applications.
Unsupervised optical flow from event cameras reaches supervised-method accuracy
A 2025 paper demonstrates unsupervised joint learning of optical flow and intensity from event cameras — eliminating the need for ground truth flow labels while reaching accuracy comparable to supervised methods.
Egocentric event cameras advance gesture recognition in first-person applications
EgoEvGesture demonstrates gesture recognition from a head-mounted event camera that handles the high ego-motion and variable lighting typical of first-person sensing — a capability that transfers to any platform where the sensor shares rigid body dynamics with the observer.
Nature Communications Engineering surveys neuromorphic computing for robotic vision
A 2025 review in Nature Communications Engineering traces the algorithmic and hardware advances driving neuromorphic sensing toward real-time robotic deployment — and identifies the remaining barriers.
Mobile computing perspective on event cameras identifies the abstraction layers that are missing
A 2025 paper frames event cameras as a mobile computing problem and identifies the abstraction, algorithm, and acceleration layers needed to make them practical on resource-constrained platforms.
ECCV 2024 workshop: combining event and frame data for more robust drone detection
A workshop paper at ECCV 2024 demonstrates that fusing event-camera and RGB frame data for drone detection outperforms either modality alone — and analyzes when each modality dominates performance.
Low-power neuromorphic virtual tripwire for drone detection at ECCV 2024
An ECCV 2024 workshop paper demonstrates a sub-100mW neuromorphic virtual tripwire for drone detection — a hardware-software co-design that runs continuously at power levels sustainable on battery-operated sensors.
IEEE survey covers neuromorphic event cameras in robotics with human-in-the-loop scenarios
A 2025 IEEE review analyzes where event cameras are making inroads in robotic systems that keep humans in the decision loop — a relevant model for semi-autonomous C-UAS applications.
Masked autoencoder pretraining advances event-based point-cloud action recognition
A 2025 paper adapts masked autoencoder pretraining to event-camera point clouds, enabling self-supervised learning that dramatically reduces the labeled data requirement — useful for domains where collecting labeled data is difficult.
SMamba applies adaptive sparsification to state-space models for event-camera processing
SMamba introduces an adaptive sparsification mechanism for Mamba state-space models processing event-camera data, achieving better accuracy-efficiency tradeoffs than previous approaches on standard event-camera benchmarks.
DEIO: deep event inertial odometry integrates events and IMU for robust state estimation
DEIO demonstrates a learned event-inertial fusion architecture that maintains accurate pose estimates during high-dynamic maneuvers where frame-based visual inertial odometry degrades — directly relevant to seeker state estimation.
ESVO2 achieves direct visual-inertial odometry with stereo event cameras
ESVO2 extends stereo event visual odometry with IMU fusion and direct depth estimation, demonstrating robust state estimation at angular rates above 300 deg/s in outdoor environments.
Event-camera optical flow reaches new efficiency frontiers with mesh-based representations
Learning Efficient Meshflow from Event Cameras demonstrates that mesh-based scene flow estimation can run at event rates with significantly lower computational cost than pixel-dense optical flow methods.
Mixture-of-experts detector and new benchmark for event-camera object detection
A 2024 paper introduces both a new large-scale event-camera detection benchmark and a mixture-of-experts detector that adapts its architecture to local event density — a dual contribution that advances both evaluation and methods.
Associative memory extends event-camera object tracking to long-duration sequences
A tracker augmented with associative memory demonstrates event-camera object tracking over sequences 10–100x longer than previous benchmarks — addressing the track continuity problem that limits most current event-camera trackers.
Neuromorphic driver distraction detection demonstrates embedded SNN deployment on event cameras
Spiking-DD demonstrates a complete spiking neural network pipeline for driver distraction detection on an event camera, running on embedded hardware — one of the first end-to-end SNN deployments validated in a real application.
Survey of event-based object detection for autonomous driving shows transfer potential to aerial sensing
A comprehensive 2024 survey of event-camera object detection in autonomous driving identifies the network architectures and dataset characteristics that transfer best to other high-speed sensing domains, including aerial target detection.
Embedded graph convolutional networks bring GNN processing to SoC FPGAs for event cameras
A 2024 paper demonstrates real-time graph neural network inference on event-camera data running on a system-on-chip FPGA — closing the gap between state-of-the-art event processing algorithms and embedded hardware targets.
State space models reshape how event cameras process temporal data — a CVPR 2024 result
Zubić et al. show that structured state space models handle the asynchronous, irregular nature of event streams better than recurrent networks or transformers, with lower latency and competitive accuracy on recognition tasks.
Fusing event cameras with infrared imaging at CVPR 2024 — what the combination unlocks
A CVPR 2024 paper demonstrates complementary fusion of event-based visible-light sensing with infrared, achieving target detection under conditions where either modality alone fails.
A comprehensive survey of event-based stereo depth estimation maps where the field stands
Arxiv 2409.17680 surveys event-based stereo depth estimation from first principles through current state-of-the-art, cataloguing methods, datasets, and the open problems that remain.
Event cameras guide low-light video enhancement at ECCV 2024 — night operations get more capable
An ECCV 2024 paper uses the event camera's high dynamic range and microsecond timing as a guide signal to enhance conventional video under low-light conditions where frame cameras degrade severely.
Frequency-aware event-guided deblurring at CVPR 2024 puts sharp frames back in reach
A CVPR 2024 method uses the event stream's frequency-domain structure to guide deblurring of conventional video frames, recovering sharp imagery from motion blur that occurs at UAS closing speeds.
Spiking neural networks for fast-moving object detection close the gap with conventional CNNs
Arxiv 2403.10677 demonstrates spiking neural network architectures for detecting fast-moving objects from event camera data, achieving accuracy competitive with ANN approaches at a fraction of the energy cost.
Stereo event-based visual-inertial odometry demonstrates robust state estimation for agile platforms
A Sensors 2025 paper presents a tightly-coupled stereo event-camera and IMU odometry system achieving reliable pose estimation at angular rates and linear accelerations that exceed what monocular or frame-based VIO can handle.
Asynchronous visual-inertial odometry for event cameras sets a new performance bar in ISPRS 2024
An ISPRS 2024 paper presents an asynchronous VIO framework that processes events without temporal binning, achieving lower drift and better robustness on fast-motion sequences than any prior event-camera VIO method.
Synthetic event camera data for spacecraft rendezvous shows domain transfer that generalizes to other terminal problems
A ScienceDirect 2025 paper generates synthetic event camera datasets for spacecraft proximity operations, demonstrating that simulation-to-real transfer is achievable and opening a data generation path for other target classes.
Space target detection with event cameras demonstrates low-latency detection against point sources in complex backgrounds
A Springer 2025 paper applies event cameras to space object detection, demonstrating that the sensor's high temporal resolution and natural response to point-source motion enables detection of fast-moving satellites that frame cameras miss.
Arxiv 2502.18490 surveys event cameras in human-centered applications — lessons that apply beyond the lab
A comprehensive survey of event-camera use in gesture recognition, gaze tracking, and human action analysis reveals algorithm and hardware patterns that generalize to any fast-motion detection problem.
Leonardo's event camera drone detection experiments — an industry perspective on what works in the field
Arxiv 2409.16099 reports Leonardo's experimental evaluation of commercial event cameras for drone detection, documenting the real-world performance limits and practical engineering challenges from an integrator's perspective.
DAGr — truly asynchronous graph neural networks for event cameras, published in Nature 2024
The DAGr paper in Nature Machine Intelligence demonstrates a graph neural network that processes each event individually without batching, achieving lower latency than any prior event-camera detection method while maintaining competitive accuracy.
From events to enhancement — a comprehensive survey maps the full event-based imaging pipeline
Arxiv 2505.05488 provides an end-to-end survey of event-based imaging from sensor physics through reconstruction, enhancement, and application, with practical guidance on pipeline design choices.
Event-based VIO for UAV indoor navigation demonstrates GPS-free precision flight that transfers to terminal guidance
A PMC 2025 paper validates event-camera visual-inertial odometry for indoor UAV navigation, demonstrating centimeter-level position accuracy in GPS-denied environments including high-speed maneuvers.
Commercial off-the-shelf event cameras for space surveillance — the Optica result that started the conversation
The Optica paper demonstrating COTS event cameras for space surveillance established event-based sensing as viable for extreme-latency-constrained detection, and is the earliest work in our review series.