Volume counts
How many vehicles pass a point, by direction and by period.
COMPUTER VISION FOR TRAFFIC STUDIES
AI video traffic counts: turning movement counts and vehicle classification from your footage.
EAGLE BS turns the video from your traffic study into counts by vehicle type, axles and movement. Artificial intelligence does the heavy lifting, a QC audit confirms the data, and you receive reports ready for the study.
- 95%+ counting accuracy
- 360 million+ vehicles classified
- Google Cloud partner
Recorded demonstration
Recorded demonstrationCounts by direction and vehicle type
Cars and trucks move in different directions, with counting lines and separate totals for each direction.
Each direction counted and split by vehicle type.
WHAT IT IS
What is an AI video traffic count?
It is a traffic count in which cameras record the roadway and a computer vision system counts the vehicles that pass, sorting each one by type, number of axles, direction and movement. The result matches what a manual count produces, with far fewer hours of review and a video record that can be checked afterwards.
In a manual count, a team watches the footage and logs every vehicle. With AI, the system follows each vehicle across the scene and classifies it on its own. People step in where they matter most: the audit that confirms the data before it reaches the report.
Count types we support
Vehicle classification counts
The same count split by vehicle type and axle configuration, including lifted axles.
Turning movement counts (TMC)
How many vehicles go through, turn right or turn left on each approach of an intersection.
License plate reference
Plate reading, where the image supports it, to help origin-destination and travel time studies.
Speed, headway, pedestrians and bicycles are evolving capabilities and are assessed case by case.
WHY EAGLE BS
Technology from a U.S. company, proven at scale in Brazil, now available in the United States.
EAGLE BS is a U.S.-based company. Our technology earned its track record at scale in vehicle surveys on Brazilian highways and urban corridors, and it is now available to the U.S. market, with a team and support here.
- 95%+
- vehicle counting accuracy
- 360 million+
- vehicles classified since 2021
- 433,000+
- hours of video processed since 2021
A proprietary AI model
EAGLE BS trained its own model on tens of thousands of vehicles, using a state-of-the-art AI architecture, to recognize vehicle types and axle configurations that generic models do not tell apart.
Varied conditions
It operates across varied recording conditions, such as night and rain. Before each study we review the camera, angle and lighting.
Google Cloud partnership
EAGLE BS is a Google Cloud Select partner. Training and high-volume processing run on Google Cloud infrastructure.
Accuracy varies with the camera, angle, lighting and traffic volume. That is why every study goes through an audit, and why a pilot on your own video shows the result for your scenario.
HOW IT WORKS
From footage to report in four steps.
You stay in charge of the study method. We handle the path from video to numbers.
Video capture
Cameras record the count locations. Video can come from fixed cameras, temporary field cameras or files you already have.
Video by location and periodAI processing
The system detects, tracks, counts and classifies each vehicle by type, axles, direction and movement.
Vehicles counted and classifiedData audit
A review step confirms the results. Cases where the AI is uncertain are checked before they enter the report.
Verified dataReports and calculations
Totals are consolidated by period, direction, movement and class, with the calculations the study calls for.
Spreadsheets and reports
SEE IT AT WORK
The count on video, as an analyst sees it.
Each demonstration shows a counting scenario. Labels and totals on screen are the output of the analysis on that video.
Paths and crossings along a road
Vehicles are tracked along the road; paths and crossings feed counts by category and direction.
Following the whole path makes it possible to count by movement, beyond a single line crossing.
Recorded demonstrationPaths and crossings along a road
Vehicles are tracked along the road; paths and crossings feed counts by category and direction.
Counting from a curb-level camera
An avenue filmed from curb level, with cars and trucks followed in both directions and flow per minute in the panel.
It shows the count from a harder angle, common in urban studies that use temporary cameras.
Recorded demonstrationCounting from a curb-level camera
An avenue filmed from curb level, with cars and trucks followed in both directions and flow per minute in the panel.
License plate reading as a reference
A road is analyzed with vehicle detection and plate readings displayed in the scene's panel.
A plate lets you relate passages at different locations, where distance, light and framing allow a reading.
Recorded demonstrationLicense plate reading as a reference
A road is analyzed with vehicle detection and plate readings displayed in the scene's panel.
Recorded technical demonstrations. Figures for each study are checked in the audit step before delivery.
WHAT YOU RECEIVE
Data ready to go into the study.
Delivery follows the format agreed at the start of the study.
Totals by interval
Volumes consolidated in short intervals, such as 15 minutes, and hour by hour, by direction and by movement.
Classification by type and axles
Motorcycles, passenger cars, light trucks, buses and trucks, with axle configuration and lifted-axle indication.
Peak hour and calculations
Peak periods identified, plus the derived calculations your study requires.
Spreadsheets and exports
Excel and CSV files, with the list of detections available for review.
Video for audit
The analyzed footage stays available to check any number in the report.
Axle counts and vehicle classes
Classification starts from the vehicle type and its axle configuration, following the standard used by Brazil's national transport agency (DNIT). The same information supports conversion to other class schemes, such as the FHWA 13-category scheme. We validate that conversion in the pilot, on video from your location.
We do not claim agency certification. Requirements from a specific agency or review process, such as count windows, intervals and raw data submission, are handled in the scope of each study.
HOW WE CAN HELP
For firms that count vehicles, and for firms that use the counts.
The same technology supports different roles in a traffic study.
Traffic data collection firms
You already record and count. Manual review is what consumes your team.
We process your footage and return verified counts. We also work as a technology partner, with your brand on the deliverable.Environmental and urban planning consultants
Many intersections, several peak periods and a review deadline.
We process many locations in parallel and deliver volumes by movement and class, with the video kept for review.Planning, modeling and regional surveys
Models need dependable volumes at many locations, and travel patterns.
Classification counts at scale and plate reading as a reference for origin-destination work, where the image allows.Engineering firms with public clients
Transportation agencies ask for raw data, a defined format and traceability.
We deliver data in the agreed format, with the audit step documented and the footage available.START WITH A TEST
Send a video and compare it with your own count.
The most direct way to evaluate us is to use footage you have already counted by hand.
You send footage
One count location, ideally with a manual count to compare against.
We process and audit
The video goes through the same four steps as a full study.
We compare results
You receive the count by class and movement and the comparison with your reference.
At the end you know what to expect in your scenario: measured accuracy, delivery format and a proposal with scope, timeline and cost.
FREQUENTLY ASKED QUESTIONS
Questions about AI traffic counts
Short answers to the questions traffic engineers and data collection firms ask most.
What is an AI video traffic count?
It is a traffic count produced from video by a computer vision system that detects, tracks and classifies each vehicle. EAGLE BS delivers volume counts, vehicle classification counts and turning movement counts this way, with a QC audit before the report.
How does AI video traffic counting work?
In four steps: cameras record the count location, AI detects, tracks and classifies each vehicle, an audit confirms the data, and reports consolidate totals by period, direction, movement and class.
How accurate are the counts?
Vehicle counting reaches over 95% accuracy. Results vary with the camera, angle, lighting and traffic volume, which is why every study goes through an audit. A pilot on your own video shows the accuracy for your scenario.
Does it work at night and in the rain?
The system operates across varied recording conditions, such as night and rain. Quality depends on the camera and the lighting at the location, which we review before the study.
Which vehicle classes do you report?
Motorcycles, passenger cars, light trucks, buses and trucks, with axle configuration and lifted-axle indication. Other class schemes are produced by conversion.
Is an axle count enough for FHWA 13-category classification?
An axle count alone is not enough, but combined with the vehicle type and the number of units (single unit, single trailer or multi-trailer) it supports the conversion. The system already records that information. We validate the FHWA conversion in the pilot, on video from your location.
Do you deliver turning movement counts in 15-minute intervals?
Yes. Totals are consolidated in short intervals, such as 15 minutes, by approach and movement, with peak periods identified. The final layout is agreed at the start of the study.
Do I need specific cameras?
Not necessarily. The system works with footage from fixed cameras, temporary cameras and existing recordings. We review resolution, angle and lighting before starting.
Where is the system used today?
In vehicle surveys in Brazil, on highways and urban corridors, with more than 360 million vehicles classified and more than 433,000 hours of video processed since 2021. EAGLE BS, a U.S.-based company, now offers the technology to the U.S. market.
How much does it cost and how long does it take?
It depends on the number of locations, the hours of footage and the classification detail. We start with a pilot and then provide a proposal with scope, timeline and cost.
SEE ALSO
Related topics.
- AI video analytics for security camerasVideo analysis on security cameras: vehicles, people, plates and protective equipment.
- Computer vision for businessHow images and video become detection, counting and alerts across different operations.
- Google Cloud partnerThe partnership behind our cloud training and processing.
EAGLE BS
Let’s count your first location.
Tell us what kind of study you run and send a recording. We will show the count, the classification and the delivery format for your scenario.