STARVIS 2 + Clear HDR
Back-illuminated 2.4 MP sensor with Clear HDR — clean image through dawn, dusk, fog and high-contrast urban scenes.
Cameras
UCMOS-D
SKU · AAD-CAM-UCMOS · STARVIS 2 Clear HDR
A compact 2.4 MP STARVIS 2 day camera with Clear HDR and a video pipeline built to be read by an onboard detection model. One camera, two interface variants — choose native digital Ethernet or a direct MIPI CSI-2 link to your compute module. Removable IR-cut filter for low-light work.
Digital Ethernet
UDP/RTSP H.264/H.265 · <50 ms
MIPI CSI-2
Direct-to-SoC · uncompressed · <10 ms
Key Advantage
STARVIS 2 + Clear HDR for dawn, dusk, fog and high-contrast urban scenes — ultra-low latency video output, built to be read by an onboard detection model.
Highlights
Core specs and the features that make UCMOS-D the daylight eye of the AAD targeting stack.
- Digital Ethernet
- MIPI CSI-2
- 2.4 MP STARVIS 2
- FOV(H) 123°
- Ethernet or MIPI
Ultra-Low Latency
Under 50 ms glass-to-stream over Ethernet — and under 10 ms on the MIPI CSI-2 variant, so the pilot and the autopilot act on what they see in real time.
AI-Ready Pipeline
Native digital output built to be read directly by an onboard detection model — no frame grabber in the loop.
Native Digital Ethernet
UDP / RTSP H.264 / H.265 over a single Ethernet link — no analog VTX and no transcoding step.
Edge-to-Edge Optics
A target the AI can't resolve is a target it can't engage — so we don't cut corners on the glass.
Most low-cost camera lenses are only sharp in the center of the frame. Toward the edges the image softens, and detection and identification models start to miss or misclassify targets that drift away from the middle of the picture.
Every signature locked — corners and center.
UCMOS-D uses a high-quality lens that holds a uniform, sharp image across the entire frame. Targets are detected and identified just as reliably at the edges as they are dead-center — which matters most in the seconds before a target leaves the field of view.
Edges soften — targets missed at the corners.
Full Specifications
Shared optical and electrical specs, with interface-specific notes for the two output options — digital Ethernet and MIPI CSI-2.
| Detector | 2.4 MP STARVIS 2 · Clear HDR |
|---|---|
| Resolution | 1920 × 1080 |
| Pixel pitch | ~3.1 µm |
| Frame rate | 60 / 30 fps |
| Focal length | 2.5 mm |
| Aperture | F/1.8 |
| FOV (H) | 123° |
| Interface | Digital Ethernet or MIPI CSI-2 |
| Output | Ethernet · UDP/RTSP H.264/H.265 — MIPI · uncompressed to SoC |
| Default IP | 192.168.144.100 (Ethernet) |
| Latency | < 50 ms (Ethernet) · < 10 ms (MIPI) |
| Power | 5–36 V · 2 W |
| Operating temp. | −20 … +60 °C |
| Dimensions | 21 × 21 × 42 mm · 40 g |
| Adjustable | Video & image parameters available for tuning — can be applied live, in flight |
Built For
A daylight sensor tuned for real-time guidance and onboard AI across strike, ISR and gimbal platforms.
FPV & Interceptor
Real-time terminal guidance
ISR & Recon
Daylight observation & ID
Fixed-Wing UAV
Long-endurance ISR
Tuned for Our Models
UCMOS-D isn't a camera that happens to feed an AI — it's built for ours. Wield-AI runs detection and identification on the live video, drawing a box and a confidence score around every target the model recognises.
Test Dataset on Request
We can provide a representative sample dataset — and a fully trained model — so you can validate detection and identification before you commit.
In-House Tonemapping & Tuning
We set the tonemapping and image tuning ourselves, calibrated to the conditions our models are trained on — not a generic factory profile.
Camera & Model, Matched
Because we build both, sensor output and model expectations line up — stronger detection and fewer missed targets out of the box.
Technical Documentation
Want the Full Picture?
Wiring diagrams, CAN and CSI pinouts, power rails, expansion options, and step-by-step setup — everything on UCMOS-D lives in our Knowledge Base.
Learn More in Our Knowledge Base →Compatible Products
Built to work as one system — camera, compute and AI on a single digital stack.
