01/Sync
Time-synchronized
Every sensor runs on one hardware clock, so camera, LiDAR, and radar frames line up to the microsecond.
Raw sensor data, captured clean.
Most data problems start at capture. Collect brings hardware time-sync, calibration, and provenance to the very first frame, so what your fleet records is ready to curate, label, and verify from day one.
Deepen·Collect
Every sensor on one shared clock, aligned frame for frame
From rig to clean dataset.
Capture quality is decided before a single frame is recorded. Collect builds sync, calibration, and provenance into the process itself.
01
Rig
Plan the sensor suite and mount calibration targets.
02
Sync
Lock every sensor to one shared hardware clock.
03
Capture
Record synchronized multi-sensor streams in the field.
04
Verify
Check calibration and coverage before data leaves the vehicle.
05
Ingest
Land clean, provenanced data ready for the stack.
Clean by design.
Four things Collect gets right at the source, so the rest of the stack starts from solid ground.
02/Calibration
Calibration-verified
Targets and checks are built into capture, so calibration holds from the first frame, not patched in later.
03/Coverage
Every sensor, one pipeline
Camera, LiDAR, radar, IMU, and GNSS captured together and ingested in the formats your tools already read.
04/Provenance
Provenance from frame zero
An auditable record of sensor, time, and location rides along on every frame, all the way through the stack.
The first step in the stack.
Collect is where verified data begins. Clean capture flows straight into curation, labeling, and validation.
01 · Collect
Collect
Capture clean, synchronized sensor data.
Every sensor. Every format.
Bring the sensors you run and the formats your tools already read. Collect handles the capture, sync, and ingest.
Sensors and formatsCameraLiDARRadarIMUGNSSROS 2MCAPnuScenesHDF5
Capture that holds up downstream.
<0.8px
Calibration
accuracy
200+
Programs
in production
Synced
Every sensor
every frame
Tracked
Provenance
from frame zero
Capture data you can build on.
Start every dataset clean, synchronized, and provenanced, ready for the rest of the stack.