The EventCV Library for Event-Based Robotic Vision

Adam D. Hines; Michael Milford; Tobias Fischer

cs.ROSep 18, 2026

The EventCV Library for Event-Based Robotic Vision

Authors: Adam D. Hines, Michael Milford, Tobias Fischer

Organizations: QUT Centre for Robotics, School of Electrical Engineering and Robotics, Queensland University of Technology, Brisbane, QLD 4000, Australia.

Abstract

Event cameras detect per-pixel brightness changes asynchronously on microsecond timescales, with high dynamic range and low power draw. These are desirable properties for robots that move fast or work in difficult lighting conditions. However, integrating an event camera into a real-world robotic pipeline still requires substantial effort: plug-and-play drivers do not exist, event streams are recorded in a variety of incompatible file formats, and most projects rely on custom research-grade code. Here, we present EventCV, an open-source and extensible Rust library with OpenCV-style Python bindings that lowers the entry barrier to working with event cameras. EventCV provides a wide range of features: denoising filters and geometric transforms, augmentations, corner detection and unsupervised feature learning, contrast-maximization motion estimation, a video-to-events simulator, and Open Neural Network Exchange (ONNX) inference for deployment in robotic stacks. EventCV integrates the Neuromorphic Drivers package, allowing an event camera stream to be processed directly in real time. No existing toolkit covers this range of operations in one package, and EventCV builds representations and decodes files 1.1x to 3.7x faster than the currently available libraries. We deploy EventCV on a Jetson Orin AGX and present three robotics case studies spanning object detection, on-device model inference, and localization. Project webpage: https://eventcv.net.

Figures & tables

Fig. 1: The EventCV ecosystem provides efficient camera and file streaming and integrates into modern machine learning and robotics workflows.

Capability	EventCV	AEDat [ 10 ]	AEStream [ 11 ]	evlib [ 12 ]	Event utils [ 13 ]	Expelliarmus [ 14 ]	Faery [ 15 ]	jAER [ 16 ]	Tonic [ 17 ]
Read/write	✓✓✓	✓	✓✓	✓	✓	✓	✓✓	✓✓	✓
Format scope	Broad I/O	AEDAT4 read	Selected formats	Selected formats	Research formats	Prophesee formats	Several formats	Several formats	Dataset parsers
Live streaming	✓✓✓	×	✓✓	×	×	×	✓✓	✓✓✓	×
Sensor control	Bias + ROI	Not applicable	Driver dependent	Not applicable	Not applicable	Not applicable	Driver dependent	Bias controls	Not applicable
Filter/transforms	✓✓✓	×	✓	✓	✓✓	✓	✓	✓✓✓	✓✓
Event processing	Geometry + filters	Not provided	Basic transforms	Lazy filtering	Augment + motion	Windowed I/O	Stream operations	Broad filter stack	Augmentation suite

TABLE I: Comparison of event-based vision libraries and open-source projects. One, two, and three check marks denote basic, substantial, and comprehensive support for a capability. A cross denotes that it is absent.

Fig. 4: Graphical interface for viewing and live filtering event streams using ecv.play .

Fig. 5: Three examples of the representations EventCV builds, from an N-ImageNet [ 23 ] stream. The top-left corner of each panel gives the time to generate it from 106,295 events over a 300 ms window on a Jetson Orin AGX.

Fig. 7: Corner and feature detection on an event slice. Left to right: raw event frame, eFAST [ 24 ] corners, the learned FEAST [ 26 ] prototype bank, and FEAST feature assignments.

Fig. 8: Contrast-maximization motion compensation in EventCV . Top: an event slice before and after compensation, boxed regions magnified. Bottom: image variance over the translational velocity search, marking the uncompensated point (circle) and the Nelder-Mead optimum (star).

Representation	EventCV	Tonic [ 17 ]	evlib [ 12 ]	Speed-up
Polarity frame (2 ch.)	23.78	136.36	67.22	2.8 $\times$
Voxel grid (5 bins)	26.50	124.44	97.23	3.7 $\times$
Time surface	21.33	59.16	80.76	2.8 $\times$

TABLE II: Representation building time (ms, lower is better) from a 1.22 M-event slice of the EVT3 recording. Speed-up is EventCV against the next-fastest library. ch.: channels.

Library	EVT3 raw	AEDAT 4.0	HDF5
	116.3 Mev	10.0 Mev	183.9 Mev
EventCV	53.8 (1.6 $\times$ )	85.4 (1.1 $\times$ )	16.3 (3.2 $\times$ )
Expelliarmus [ 14 ]	33.5	—	—
dv-processing [ 8 ]	—	77.0	—
evlib [ 12 ]	5.4	—	5.1

TABLE III: Decode throughput (Mev/s, higher is better), single-threaded, for every reader able to open each format. A dash marks a format the reader cannot open. For EventCV , the speed-up over the next-fastest reader for each format is shown in parentheses.

Fig. 9: Latency of event counting against slice size, comparing EventCV with Tonic [ 17 ] , event_utils [ 13 ] , and evlib [ 12 ] on a Jetson Orin AGX. Both axes are logarithmic.

Fig. 10: The ev-ultralytics YOLO26 [ 41 , 42 ] pipeline on an EVK4 recording, run on a Jetson Orin AGX. Top: per-window processing time by stage, OpenEB against EventCV . Bottom: example detections from the EventCV pipeline.

Fig. 11: Capture rig for the SuperEvent [ 35 ] case study. A Prophesee EVK4 views an outdoor walking sequence replayed on a laptop screen. Events stream through EventCV into SuperEvent on a Jetson Orin AGX.

Fig. 12: Comparison of EventCV deployed on CPU with custom CUDA data formatting in Event-LAB [ 43 ] . Throughput in Mev/s, higher is better.

Event-Based Vision

Event-based vision uses asynchronous camera measurements to perceive motion and scene changes with high temporal resolution and dynamic range.

See all 232 papers

EVIS: Real-Time Event Camera Simulation with Multimodal Supervision in NVIDIA Isaac Sim
Jul 9, 2026Linli Shi, Ruijun Zhang, Ziyun WangEvent-Based Vision Robotics Simulation
Johns Hopkins University
Efficient Event Camera Volume System
Mar 16, 2026Juan Camilo Soto, Ian Noronha, Saru Bharti +1Event-Based Vision
Purdue University, USA.
EventGeM: Global-to-Local Feature Matching for Event-Based Visual Place Recognition
Mar 6, 2026Adam D. Hines, Gokul B. Nair, Nicolás Marticorena +2Visual Place Recognition Event-Based Vision
QUT Centre for Robotics, School of Electrical Engineering and Robotics, Queensland University of Technology, Brisbane, QLD 4000, Australia.

Jul 9, 2026J/K move · Enter open · S save

EVIS: Real-Time Event Camera Simulation with Multimodal Supervision in NVIDIA Isaac Sim

Linli Shi, Ruijun Zhang, Ziyun Wang

Johns Hopkins University

Event-Based Vision Robotics Simulation Synthetic Data Generation for Robotics

Open paper AskPDF arXiv

Event cameras are increasingly adopted in embodied perception for their microsecond temporal resolution, high dynamic range, and resilience to motion blur. However, training event-based models for robotics requires large-scale, action-conditioned data with dense physical annotations that are difficult to collect in the real world. We introduce EVIS, an open-source physics-grounded event simulator integrated into NVIDIA Isaac Sim that generates events from linear-HDR radiance from a closed-loop robot training episode. Rather than relying on learning-based video interpolation or expensive dense rendering, EVIS exploits renderer-provided motion vectors and depth maps through bi-directional warping and depth-based splatting. This enables high throughput, real-time, and high-fidelity event generation. We evaluate EVIS across runtime efficiency, sim-to-real transfer, and zero-shot model compatibility. A rotation-speed estimator trained solely on EVIS events achieves 2.75 rad/s MAE on real sensor data. Pretrained models for reconstruction, matching, and tracking perform competitively on EVIS events without any fine-tuning. EVIS sustains real-time generation across GPU-parallel environments on a single GPU. Code repository: https://github.com/spikelab-jhu/isaac-sim-event-camera-plugin.

Abstract

Event-Based Vision

Explore similar work

EVIS: Real-Time Event Camera Simulation with Multimodal Supervision in NVIDIA Isaac Sim

Efficient Event Camera Volume System

EventGeM: Global-to-Local Feature Matching for Event-Based Visual Place Recognition