jAER is an open-source desktop application designed for real-time neuromorphic event sensor processing and silicon cochleas. It captures hardware output over USB, runs low-latency event filtering and denoising algorithms, and routes event streams to external deep neural networks, OpenCV pipelines, and ROS2 robot controllers.
Technical Overview of jAER
jAER operates via binary desktop installers that bundle an Eclipse Temurin Java runtime, though it also runs directly from source using Apache Ant. Hardware event streaming is handled at high throughput via native libusb drivers and shared memory buffers rather than slow network sockets. Because the architecture streams raw AER data directly to memory, it processes heavy sensor feeds without external background services.
Core Features of jAER
- Broad hardware support accommodates DVS, DAVIS, and IMX636 event cameras from iniVation, Prophesee, and NRV without vendor lock-in.
- Strict local privacy guarantees zero telemetry, phoning home only during manual update checks or user-initiated data downloads.
- Low-overhead pipeline uses
DNNOutputViaSharedMemoryand fast AEDAT-4 compression to feed machine learning models without dropped events. - Extensible filter architecture automatically generates functional property panels for custom Java event filters and noise-reduction routines.
- Flexible data export converts live or recorded sensor activity directly into standard MP4 videos, CSV tables, or HDF5 containers.
System Compatibility & Updates
Binaries run on 64-bit Windows, Linux (unsandboxed for raw USB device access), and macOS, including Apple Silicon hardware via Homebrew libusb. The current release cycle targets JDK 25 and introduces verified live streaming support for Samsung S5KRC1S sensors via the DELTA01 integration.
Why Choose jAER?
It delivers twenty years of neuromorphic research into a single open-source workbench that unifies diverse, competing event-camera brands under one standard interface. Researchers gain unmetered, telemetry-free control over raw spike streams without depending on proprietary, closed-source manufacturer SDKs.