The Evolution of In-Cabin Sensing: Why OEMs are Shifting from Driver to Occupant Monitoring
Key Takeaways:
DMS were designed to watch the driver. OMS monitor the driver and its passengers.
Regulatory pressure, EV Cabin experience competition, and safety expectations are driving the shift from DMS to OMS right now - not in the future
Expanding from 1 occupant to 5 requires a fundamentally different technical approach, not an upgrade of existing systems
Re-Identification - knowing consistently who is where throughout a journey - is what makes OMS actually useful in production
Processing happens locally on the vehicle. No images, no audio, no data leave the car.
Driver monitoring did exactly what it said - to watch the driver. It checked for attention, drowsiness. It first appeared in passenger vehicles in 2006, introduced by Lexus. These early systems focused primarily on detecting driver drowsiness using infrared sensors and it helped prevent accidents caused by the one person whose attention the vehicle depended on the most.
But that was the right problem to solve at the time.
And the problem has now moved.
Today’s vehicles are smarter, more automated, and increasingly designed around the experience of every person inside - not just the one behind the wheel. And the monitoring systems inside them need to catch up. That is what the shift from Driver Monitoring System (DMS) to Occupant Monitoring System (OMS) is: not a product upgrade, but a critical rethink of what cabin intelligence is actually for.
DMS now, OMS later?
Driver monitoring systems emerged in response to a specific, well-defined problem: fatigue and distraction cause accidents. 90% of serious crashes linked to human error, and driver fatigue is the responsible cause of road crashes in the UK. The solution was to instrument the driver - track gaze, detect drowsiness, monitor head position - and intervene before something went wrong.
DMS are not yet in every car on the road, but they are rapidly becoming mandatory standard equipment for all new cars in major markets, particularly Europe, by July 2026 to comply with the EU General Safety Regulation (GSR) 2019/2144.
But not for long. OMS are also gaining traction, driven by safety regulations requiring child presence detection. Would DMS be enough? It only assumes one occupant mattered. It assumes that occupant was in the front left seat and it assumes the vehicle’s job was to just keep the person alert, not to understand anything more complex about the humans inside the vehicle that could also affect the driver’s journey.
The swift changes
This regulation expands to the camera infrastructure. Mandating DMS means cabin-facing cameras and the compute to run them became standard fitment. Once that hardware was in the vehicles, the question becomes: what else should it do? Does one camera mean for just one driver? But what about the occupants?
EVs changed what vehicles compete on. The differentiation narrows in the electric vehicle segment, the interior becomes the product. Ambient lighting, personalised climate, adaptive sound, intelligence response to how occupants feel - these are active battlegrounds between manufacturers. A system that understands only the driver can only personalise the driver’s experience. What about the passengers?
Passenger safety expectations are also growing. Child presence detection, rear occupant alerts, seatbelt compliance, distress detection in vulnerable passengers (read Euro NCAP 2030 here) - none of them are addressable by a driver-only system. As vehicles become more automated and journeys longer, the expectations that the vehicle is aware of everyone inside it has shifted from a luxury feature to a basic safety assumption.
Let’s address it - OMS is not DMS with more cameras
Multi-person cabin monitoring is now considered an essential, safety-critical requirement in the automotive industry, rather than a luxury or optional feature. This shift is driven by stringent regulatory mandates, advances in autonomous driving, and the need to enhance overall passenger safety. This is the most critical point to understand about the shift - and the most commonly misunderstood.
Occupant monitoring is not driver monitoring scaled up. Processing X number of occupants simultaneously in real-time, across front and rear seats, in variable lighting, with natural head movement and partial occlusion, at the latency required for production automotive applications - this requires a model architecture built for the problem from the ground up.
Re-Identification - the capability that makes it real
Understanding that there are five people in a vehicle is the easy part. Knowing consistently which person is which - across an entire journey, as they move, turn, sleep, and shift position - is what re-identification solves.
Without it, an occupant monitoring system is not actually monitoring occupants.
Re-identification maintains a continuous, accurate model of each occupant from the moment they enter the vehicle to the moment they leave. It handles occlusion - a passenger turning away, wearing glasses, a mask, a child obscured by a headrest. It handles re-entry - someone who gets out at a service station and returns. It is the connective tissue that turns a sensing capability into a coherent system.
And critically, in our implementation, it happens entirely on the vehicle. No images are transmitted. No biometric data leaves the cabin. Re-identification runs on-edge, which means it works regardless of connectivity and complies with data privacy requirements by architecture, not by policy.
What this means for your roadmap ahead
OMS is not an up-and-coming technology or capability. It is in the requirements and regulatory programmes that will reach production within the next few years. The OEMs and Tier 1 suppliers building for it now are not early adopters- they are on schedule.
The ones treating it as a future consideration are not.
The cabin has always carried more than one person. The systems inside it are finally starting to reflect that.
The bottom line here is…
Driver monitoring solves a a critical problem. Occupant monitoring solves the complete problem. The shift is being driven by compliance and regulation, by market competition and by the reality that modern vehicles are shared spaces that deserve shared awareness. The technology to do this properly - accurately, in real-time, on embedded hardware, without sending personal data anywhere - exists today. The question here is whether they are simply integrating it to their systems.
Frequently Asked Questions
What is the difference between DMS and OMS?
Driver Monitoring Systems (DMS) focus exclusively on the person behind the wheel. Using cameras and infrared sensors, DMS tracks driver’s eye movement, head position, and blinking patterns in real-time to detect signs of drowsiness, distraction, or inattentiveness, issuing safety alerts to prevent accidents.
Occupant Monitoring Systems (OMS) expand this field of view to the entire cabin. By monitoring all passengers - front and rear - OMS enables a broader range of applications. While DMS is primarily about preventative safety, OMS unlocks personalidation, passenger wellbeing, and interior intelligence for every seat in the car.
Is occupant monitoring required by regulation?
Yes, the regulatory landscape is shifting quickly. While Driver Monitoring is already mandated in Europe under General Safety Regulation 2 (GSR2), Occupant Monitoring is becoming a critical differentiator for safety ratings. Euro NCAP scoring now includes criteria for Child Presence Detection (CPD), and emerging OEM specifications are increasingly requiring rear-seat awareness to meet five-star safety standards.
Does re-identification work without sending data to the cloud?
BLUESKEYE utilises a "Privacy by Design" architecture. Re-identification - recognising a returning occupant, runs on edge and ties IDs to identifiable biomarker traits per ID which happens entirely on-edge within the vehicle’s embedded hardware. No raw images or personal voice recordings are ever stored or transmitted to the cloud, ensuring total user privacy and compliance with the EU AI Act.
Does your occupant monitoring work in low light or with occlusions?
Yes - our models are specifically trained to work with both RGB (standard color) and NIR (Near-Infrared) cameras. This ensures 24/7 performance, even in low light.
Regarding occlusions (such as glasses, face masks, or hats), our SDK uses robust landmark tracking that can infer behaviour and states even when parts of the face are partially obscured. While total occlusion - where the entire face is obscured or occluded - naturally limits face tracking, our system is engineered for maximum resilience.
What hardware does the B-Automotive SDK run on?
The B-Automotive SDK is hardware-agnostic and highly optimised for the automotive supply chain. It runs seamlessly on:
Standard CPUs: ARM (v7, v8, and v9 architectures) and x86.
Specialised chipsets: We offer proven support for the Samsung Exynos and have successfully deployed on NVIDIA Jetson platforms.
Custom integration: We can provide custom GPU and NPU (Neural Processing Unit) acceleration to minimise CPU load.
For full performance specifications, including memory footprint and inference latency, email us at sales@blueskeye.com