The Smartest Traffic
Management Solution.
Currux Vision is the most advanced AI-based detection and actuation platform built for a future-proof city — one fully integrated hardware-and-software system engineered to see every vehicle, pedestrian and cyclist, in real time, in any condition, and act on what it sees.
UNIT — GEN 3 PLUS
STATUS — FIELD READY
Independently tested. Publicly documented.
Corridor delay and emissions cut within one week of activating Adaptive Corridor Control across a county of nearly one million residents.
Read the case study →Detection accuracy confirmed by the San José DOT across day, night, rain, camera vibration and partial obstruction.
Read the case study →Accuracy independently verified in joint testing of the integrated Currux Vision and Costar traffic video analytics platform.
Read the case study →One system. Not a stack of vendors bolted together.
Most agencies run their intersections on a patchwork — loops from one era, radar from another, video detection from a third party, analytics from a fourth. Currux Vision replaces the patchwork with a single, fully integrated AI hardware-and-software platform: detection, actuation, digital twin, adaptive timing and fleet intelligence, engineered end-to-end and manufactured by the same company that writes the software.
Legacy Detection
- Buried loops and radar pucks — blind to pedestrians and bicycles, expensive to replace, invisible until they fail.
- First-generation video detection loses accuracy in shadows, glare, rain, snow, fog, low light and pole sway.
- Detection, analytics and actuation sourced from different vendors, integrated by whoever's left holding the ticket.
Currux Vision
- One camera-agnostic AI system — vehicles, pedestrians and bicyclists, counted and classified in the open and in the crosswalk.
- Trained and hardened for the conditions that break legacy systems, holding 99.6% validated detection accuracy across all conditions — day, night and in weather.
- Hardware, software, digital twin and AI operations tools engineered by one team, in Houston, Texas — no third-party black box.
Built to see through anything.
Shadows. Glare. Rain, snow, fog, low light — even a camera pole moving in the wind. Legacy video and radar lose the plot exactly when it matters most. Currux Vision's models are trained on conditions like these, not just clear-day footage, so detection holds steady around the clock.
Hardware we don't outsource.
The Gen 3 Plus AI Edge Server is engineered entirely in-house, from the carrier board up — not a third-party appliance with our logo on it. That means tighter quality control, faster root-cause fixes, and a roadmap nobody else controls.
- NVIDIA Jetson AGX Orin — 8-core ARM CPU, the same chip family trusted by Rivian and Toyota for autonomous systems.
- Cabinet-ready — 4.0"H × 6.8"L × 6.7"W, DIN rail mount, NEMA TS2 Type 2 compliant.
- Built-in switching — 4× 1GbE (3-port integrated switch + 1 dedicated uplink), cleaner cabinet networking, no external switch.
- Engineered for the field — 12V PWM closed-loop cooling, RTC holdover battery, at-a-glance SDLC status LEDs, optional 20 kA surge protection.
- Up to 4 cameras per server, 8 licensed camera slots, NVMe-first boot for speed and longevity.
| Processor | NVIDIA Jetson AGX Orin SOM — 8-core ARM Cortex v8.2, up to 2.2 GHz |
| Power | 9–20 VDC, ≤80 W |
| Memory / Storage | 32 GB LPDDR5 @ 3200 MHz · ≥120 GB PCIe Gen3 NVMe SSD |
| Cooling | 12V PWM-controlled fan with tachometer feedback + heatsink |
| Dimensions | 4.0"(H) × 6.8"(L) × 6.7"(W) — DIN rail mount |
| Standard | NEMA TS2 Type 2 |
| Connectivity | 4× 1GbE · 2× USB 3.0 · HDMI 2.1 (4K@60Hz) · SDLC · TS1 Relay · 5IO Relay |
| Options | 4-channel analog camera input · SDLC surge protection · built-in Verizon-certified 4G modem |
| Made | Engineered, manufactured and serviced in Houston, Texas — typical repair turnaround 1–2 weeks |
Watch your city think.
Every AI detection is mirrored, live, into a real-time graphical twin of the intersection — one screen for every approach, every camera, every phase. No separate system to stand up: the twin starts the moment Currux Vision's AI processing goes active.
- Real-time counts, speed, arrivals on green, level of service and queue length, by lane and by class.
- Historical analytics with heat-map overlays by phase and approach — volume, delay, wait time, queue and controller timings.
- Controller-aware — where an NTCIP link exists, the twin shows live max-green times alongside Currux Vision's recommended adaptive timings.
DIGITAL TWIN — LIVE
Every server reports its own AI, communications and controller status — searchable by name, filterable by agency, plotted on one map instead of one browser tab per intersection.
| Live Video Wall | Every intersection's feed, side by side, from one screen. |
| Compare Mode | Select up to four intersections for side-by-side review. |
| Event Feed | One chronological stream of notifications across the fleet. |
| Alert Rules | e.g. trigger when more than 5 wrong-way detections occur in 15 minutes. |
Every intersection, one screen.
Currux Vision's Cloud Monitoring Tool puts every edge server, every Digital Twin and every camera feed into a single, centralized interface — reachable from anywhere, without a site visit. Layered on top, the Virtual TMC turns that same fleet into a live command view: video wall, event feed and alerting in one place.
- One map, whole fleet — every server discoverable by name or agency, with live status for AI processing, communications and controller links at a glance.
- Cloud or offline — cloud-hosted for servers with internet access, or a discovery-based desktop application for networks that keep units off the public internet entirely.
- Built for a real TMC — a live video wall, side-by-side comparison, and one unified event feed replace a browser tab per intersection.
- Alerts tuned to the agency — set thresholds by event type and frequency, not just a single trigger per camera.
Your fleet never sleeps. Neither does Sophia.
Sophia AI is Currux Vision's AI operations analyst — the layer above the dashboard that actually tells engineers what to look at, before a camera fails or a corridor quietly gets worse.
Four metrics — Detection, Lane Alignment, Video Corruption and Abnormal Traffic/Queueing — roll up into two clear scores per camera: a Unit Operation Score for the health of the camera itself, and a Road Score for what's happening on the road it's watching.
A low score is a prompt to look at the camera behind it — every score links back to the exact image and moment that produced it. Nothing is a black box.
A daily briefing, not a wall of alerts.
A single busy intersection can generate thousands of events a day. Sophia's Daily Briefing removes the noise: one short, automated read every morning of what actually changed across the whole agency — ready before the first coffee, on one page.
- Ordered like an engineer triages — equipment health first, then what the fleet detected, then corridor performance.
- Measured against its own normal — every event is compared to that intersection's recent history, so only a real departure gets raised.
- Remembers yesterday — items are marked new, ongoing, escalating or resolved, so a long-running issue never reads as news.
- Names the intersection driving the numbers instead of averaging it into the rest of the fleet.
| Wait Time | How long vehicles sit stationary at a red before it turns green. |
| Arrivals on Green | Share of vehicles that hit a green light instead of stopping. |
| Saturated Flow | Maximum sustainable vehicle throughput during the green interval. |
| Total Delay | Cumulative time every vehicle spends waiting, corridor-wide. |
| Time to Clear the Queue | How fast a full queue is served once the light turns green. |
| CO2 Emissions Avoided | Corridor-wide emissions saved versus the previous static timing plan. |
Every metric is measured before and after adaptive timing goes live — intersection by intersection, and corridor-wide.
Traffic that times itself.
Fixed time-of-day plans run on a schedule, not on what's actually happening at the intersection. Currux Vision's Adaptive Module continuously blends historical traffic patterns with real-time AI detection to retime a corridor while traffic is moving through it — no new wires in the ground, no rip-and-replace. It's one of the fastest levers a city has for cutting delay and congestion without a construction budget.
- Hybrid timing strategy — merges a 14-day historical baseline with live detection, recalculating every few seconds instead of waiting on a fixed cycle.
- Corridor coordination — intersections share real-time arrival data to build smooth green bands, extending mainline green and suppressing low-priority side-street calls as vehicles approach.
- Pedestrian-aware — timing can react directly to pedestrian volume, or favor pedestrian movements through a configurable priority weighting inside the adaptive logic itself.
- Scales with the agency — preview proposed changes before they ever reach the controller, then grow from one intersection to a full corridor at your own pace.
- Fails safe — if live detection or corridor connectivity drops, the system automatically falls back to the historical timing plan. Never stuck, never blind.
Zero is the only acceptable number.
Currux Vision was built around Vision Zero from the start. Every camera watches for the moments that matter — a pedestrian entering a crosswalk against the signal, a red-light runner, a near miss measured in fractions of a second — and can act immediately: placing a call, triggering a roadside-unit warning to a connected vehicle, or firing a relay to a flashing beacon.
- Conflict & near-miss analytics — Time-to-Collision and Post-Encroachment Time measured for every road user, with heat maps by conflict area.
- Violation detection — red-light running, crosswalk violations, wrong-way and speeding events, each with photo/video evidence.
- Connected-vehicle actuation — real-time object data streamed to roadside units (RSU) for pedestrian-in-crosswalk and red-light-running warnings.
- Best-in-class pedestrian & bicycle detection, in the open road and inside the crosswalk itself.
One platform, built out further than anyone needs on day one.
Everything below runs on the same camera, the same server, and the same software license — deployed as a city's needs grow.
Detection & Actuation
NTCIP/SDLC vehicle and pedestrian calls by class, with extension logic for bikes and pedestrians in the crosswalk.
Digital Twin
Real-time 3D view of every monitored approach, with exportable historical analytics.
Adaptive Timing
Historical and real-time optimization from a single intersection to a full corridor.
Sophia AI
Continuous camera health scoring and a daily briefing that flags what changed.
Vision Zero Analytics
Conflict, near-miss and violation detection with photo and video evidence.
Train Detection
Classifies trains, counts wagons, and tracks direction, speed and blocked-crossing time.
Connected Vehicle / RSU
Streams real-time object data to roadside units for V2X safety applications.
Central & Cloud Monitoring
One screen for every server's health, live video and Digital Twins, on or off the public internet.
Garage Occupancy
Real-time entry/exit counting with API and contact-closure outputs for third-party signage.
AI+ Model Training
Continuous retraining for a specific intersection, sustaining 99.6% validated detection accuracy as seasonal and site conditions shift.
The platform, on camera.
Five short films from the field and from Currux Vision's Houston headquarters — the hardware on the pole, the AI reading a foggy intersection, and the story behind the platform.
Pioneering the Future of Traffic Management
An inside look at the integrated hardware-and-software platform setting the pace for the next generation of traffic management.
Watch on YouTube ↗Gen 3 Plus
A close-up look at the Gen 3 Plus AI Edge Server — the NEMA TS2-compliant hardware engineered and built in Houston, Texas.
Watch on YouTube ↗Unveiling the Power of Camera-Based AI in Any Conditions
Camera-based detection put up against the conditions that break legacy sensors — rain, fog, glare and low light.
Watch on YouTube ↗Building Smarter Cities with AI from Houston's Heart
The Currux Vision story — how a Houston-based engineering team is building the AI backbone for smarter, safer cities.
Watch on YouTube ↗Transforming Cities, One Journey at a Time
How Currux Vision's AI mobility platform is reshaping everyday streets into intelligent, self-aware infrastructure.
Watch on YouTube ↗
Designed, engineered and built in Houston, Texas.
Every Currux Vision server is designed, integrated and assembled in-house in Texas — eliminating third-party appliances, the integration headaches that come with them, and the wait for someone else's fix.
Straight answers for the people who spec this.
Hardware and software that gives roads, intersections and highways the ability to sense and respond to real-world traffic — vehicles, pedestrians, cyclists — using computer vision and edge AI instead of buried loops or radar. Currux Vision builds the full stack: camera-agnostic detection, purpose-built edge hardware, real-time actuation, and AI-driven operations tools like Sophia AI and the Digital Twin.
99.6% validated detection accuracy across all conditions, with best-in-class pedestrian and bicycle detection out of the box. Agencies enrolled in the AI+ Model Training service get continuous retraining for their specific intersection, up to three times a year, to sustain that accuracy as conditions change.
Yes. Currux Vision is camera agnostic — it works with existing IP and analog cameras as well as new fisheye cameras, so agencies don't need to rip and replace infrastructure to get AI-based detection.
Yes. The Gen 3 Plus AI Edge Server is engineered, manufactured and serviced in Houston, Texas — design, manufacturing and assembly are handled in-house, which is why units are typically repaired in one to two weeks with local warranty support, and why the platform is Buy America / Build America compliant.
Currux Vision's AI operations analyst. Sophia AI Standard scores every camera roughly every 15 minutes on a 1–5 scale across detection, lane alignment and video health, and Sophia's Daily Briefing delivers one short automated read each morning of what changed across an agency's fleet.
A real-time graphical view of every monitored intersection, built live from camera AI detections — counts, speed, arrivals on green, level of service and queue length — with historical analytics and heat-map overlays by phase and approach.
Yes. Currux Vision communicates with NEMA-compliant traffic controllers over NTCIP and SDLC, places vehicle and pedestrian calls by category, and can actuate directly through an onboard I/O relay module, from a compact NEMA TS2-compliant server designed to install directly in the cabinet.
Ready to give your city sight?
Tell us about your corridor, your cabinet, or your camera inventory — we'll show you exactly how Currux Vision fits.