OMNIVISION OAX7700 is a newly launched ultralow-power automotive vision processor designed to add always-on parking surveillance to existing exterior camera systems. Samples are available now and mass production is scheduled for the first quarter of 2027, making the launch a component roadmap rather than a claim that cars already ship with it.
Key takeaways
- The OAX7700 combines an NPU with stacked memory in a 5.5mm by 3.5mm package.
- It is designed to keep parking-surveillance vision active while conserving vehicle power.
- Local detection can trigger recording on confirmed people or objects instead of every movement.
What the OMNIVISION OAX7700 does
The chip sits beside an automotive image sensor and processes the camera stream locally. OMNIVISION says its neural processing unit can support human-presence detection, object classification and distance estimation. The design targets surround-view camera systems that already look around the vehicle but normally power down or use simpler motion triggers while parked.
Automotive World independently described the part as an AI parking-watch processor for vehicle cameras. The key qualification is that the algorithms and production integration remain the responsibility of automakers and suppliers; the release announces silicon capability and availability, not a finished consumer safety system.
Why low power is the actual product
A parked vehicle cannot treat its 12-volt battery like an unlimited data-centre supply. Keeping several cameras and a general-purpose compute stack awake would consume energy and create heat. OMNIVISION’s pitch is to move a narrow inference task into a small, dedicated processor so the system can watch continuously and wake heavier recording or vehicle systems only after a relevant detection.
That selective trigger is the commercial mechanism. A camera can ignore harmless background motion, retain a short pre-roll around an incident and preserve evidence without continuously running the full infotainment computer. The value therefore depends less on headline AI performance and more on power budget, false-trigger control and integration with cameras already qualified for automotive use.
What manufacturers still have to prove
The OAX7700 supports over-the-air feature updates and customer algorithms, but those capabilities also create validation work. Vehicle makers must define what counts as a threat object, test edge cases across weather and lighting, secure the update chain and explain retention rules for recorded footage. As Plugin4Shell showed, a deployment claim is only as strong as its verification controls.
The hardware timeline matters too. Samples allow engineering teams to prototype now, while first-quarter 2027 mass production sets the earliest broad supply window. A production car programme will still need design, functional-safety, privacy and regional compliance decisions.
The practical consequence
The quotable bottom line: OMNIVISION OAX7700 turns parking surveillance into an edge-computing problem: detect locally at very low power, wake the recorder selectively, and keep the main vehicle computer asleep.
That architecture can make evidence capture more useful, but it is not a guarantee of security or safety. Buyers should look for the automaker’s final camera coverage, retention settings, battery-impact testing and update policy rather than treating the presence of an AI processor as the finished feature.
Facts at a glance
| Item | Verified detail |
|---|---|
| Disclosure | September 21, 2026 |
| Package | 5.5mm × 3.5mm, 60-pin BGA |
| Functions | Presence, object and distance detection |
| Production | Planned for Q1 2027 |
Related Lapaas Voice coverage: related technology coverage and related technology coverage.
Frequently asked questions
What is the OMNIVISION OAX7700?
It is an ultralow-power automotive vision ASIC with an NPU for always-on parking surveillance and local object detection.
Is the OAX7700 available now?
Engineering samples are available; OMNIVISION plans mass production in the first quarter of 2027.
Does it record everything around a parked car?
The processor is designed to classify relevant objects and trigger recording selectively, but final behaviour depends on the vehicle maker’s implementation.
Sources
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
