Robotics sensor-data stack
apairo is an open, modular toolkit for robotics sensor data — one numpy-native API to load, transform, preprocess and visualise LiDAR, images and trajectories across datasets.
pip install apairo fig.01 — the pipeline, end to end
01 / The stack
Each package does one thing well and snaps onto the next. Start anywhere — the apairo core is the only hard dependency.
02 / Packages
Pick the layers you need. They share one data model, so they always agree on what a frame is.
Core · dataset loader
Unified Python loader for robotics sensor datasets — one API across synchronous and asynchronous layouts. Everything comes back as plain numpy arrays.
pip install apairo Ingest
Turn ROS bags into an apairo-native dataset — interactive TUI, KITTI-style .apairo layout, /tf handling and per-channel frame ids.
Runtime
Transforms & augmentations applied lazily at access time — nothing written to disk, numpy-only, no heavy dependencies.
Offline
Offline pipelines for LiDAR odometry, ground segmentation and traversability ground-truth generation.
Visualise
Rerun visualisation layer for clouds, semantic labels and trajectories — side-by-side pipelines, sequence-aware navigation.
03 / Quickstart
The core ships on PyPI. Add only the layers you need — every package speaks the same numpy-native data model, so nothing has to be glued together by hand.
import apairo
from apairo_transform import RangeFilter, RandomSubsample
# 1 · Load — one API across every dataset
ds = apairo.Rellis3DDataset(
"/data/RELLIS-3D",
keys=["lidar", "labels"],
)
# 2 · Transform — lazy, on access, nothing written to disk
ds.transform("lidar", RangeFilter(max=50.0))
ds.transform("lidar", RandomSubsample(8192))
# 3 · Use — plain numpy, ready for any framework
sample = ds[0]
sample.data["lidar"] # ndarray (N, 4) · x y z intensity
sample.data["labels"] # ndarray (N,) · int64