To drive itself, a vehicle must first learn to see.
An autonomous vehicle is, first and foremost, a camera that understands what it sees. There are approaches that stack sensors — LiDAR, radar, ultrasound — each costing tens of thousands of dollars. And there is the approach that bets on computer vision: cameras and artificial intelligence.
The Bex Cam is our bet on the second school — applied to the Brazilian context. Two products, two purposes: one to scale the collection network, another to train the models that will drive.
Bex Cam App
The most direct path to putting eyes on Brazil’s streets. Any smartphone becomes a collection station — video, GPS, accelerometer — while the driver drives normally.
The App’s purpose is not to train driving models. It is coverage. Thousands of drivers in dozens of cities, generating real-time traffic data. Every phone connected is one more node in the Malha Bex — the network that turns Brazil’s vehicle fleet into an urban monitoring infrastructure for municipalities, fleets, and insurers.
Bex Cam Device
A dedicated camera, installed in the vehicle, with full OBD telemetry. Resolution, angle, frame rate, and metadata — all controlled to generate training-quality data.
The purpose is different from the App: every kilometer driven with the Device feeds the Bex Dataset and trains the Bex Pilot. The Device does not seek scale — it seeks precision. It is the sensor that teaches the model to see motorcycle taxis, potholes, and wrong-way overtaking on two-lane roads.