NeoEyes NE302 Compact STM32N6 AI Vision Node

Product overview

NE302 is built for product teams that need to determine what happened in a fixed scene or what state equipment is in, then pass that result to another system. It combines image capture, task-specific inference, device logic, local records and network delivery in one compact node. Teams can validate the workflow with the complete two-board device, then use the Main Board path when designing NE302 into their own product.

Around the STM32N6, NE302 adds a device-level deployment workflow with a Web console, triggers, records, retry policies and result-delivery protocols. The NE302 Quick Guide shows the path from live-image inspection to the first local result. For fixed-scene state and event decisions, the MCU-based design reduces the space, power and software overhead of a Linux multi-TOPS platform. Projects that require multiple image sensors, depth, robotic perception, large models or general video surveillance need a different class of system.

Compatibility Notice

Key Features

From evaluation to product integration

The complete device combines a Main Board and Interface Board. It provides USB Type-C power, MicroSD, controls and service connections for evaluation and deployment. Teams can later use the 38 x 38 mm Main Board with project-specific DC power inside their own product. The Interface Board MicroSD slot does not apply in that form. Check the NE302 component overview before planning connectors, storage or an enclosure.

Validate the image-to-result path

The Web console lets a team inspect the live image, capture a known frame, load a packaged model and review its JSON result. This controlled sequence helps identify whether a problem comes from the scene, lens or model before the team connects an external application. The NE302 model validation guide covers package requirements, fixed-image testing and result inspection.

Event-driven capture and low-power control

The STM32U0 always-on control domain supports low-power sleep and multi-source wake-up. NE302 can capture from PIR/IO, a schedule or remote MQTT, then upload the result, retain it for retry or keep a local record in internal Flash or on MicroSD. The capture and storage guide explains trigger settings, storage choices, retention and record review.

Local decisions with connected results

The STM32N6 runs INT8 inference on the device, so an application can receive structured results without sending every source image to a cloud service. The current firmware reports through MQTT/MQTTS or Webhook. RTSP service and RTMP push are available when a workflow needs media during setup or operation. Use the data transmission guide after the local result is correct and the next task is connecting it to an application.

Source access for product work

The engineering project separates FSBL, App, Web, Model and WakeCore components. Source access supports firmware and product integration work after the standard device workflow has been validated. N6 firmware and the U0 WakeCore use different programming paths, so confirm the delivered board and adapter before flashing. The build and flash guide documents the supported workflow.

Lens selection

Wi-Fi HaLow version expected in late September 2026.

Both WiFi HaLow SKUs include 868 MHz and 915 MHz . Lens comparisons assume the same mounting distance. Verify wireless coverage and model recognition at the final installation.

What’s included

• 1x NeoEyes NE302
• 1x antenna matched to the ordered SKU
• 1x USB Type-C cable
• 1x 3M adhesive