Tutorials

These tutorials go from a first connection to a recorded dataset. Each page explains one script in PythonAPI/examples/tutorials/ and shows what it writes and prints. Every image and transcript on these pages comes from running the scripts against the release server at its default resolution (1280 × 720).

Tutorial and script	Covers
First steps first_steps.py	connecting, the spectator and streaming, synchronous stepping, time of day and weather
Cameras cameras.py	a vehicle on autopilot, attached RGB cameras, saving frames
Semantic and instance segmentation segmentation.py	label cameras, the class palette, instance ids and masks
Depth depth.py	metric depth, 16-bit PNG, a colour-mapped preview, back-projection to a point cloud
Bounding boxes bounding_boxes.py	visible and amodal 2D boxes, occlusion, 3D poses, overlays, per-frame JSON
Recording a dataset record_dataset.py	traffic and pedestrians, a synchronized sensor rig, a COCO annotation file
Controlling actors actors.py	manual driving, autopilot with a route, walkers, the AI controller, destroying actors

Before you start

Start the server as described in Getting started and leave it running. Install the Python package and numpy, which every tutorial from segmentation on uses for arrays and drawing:

pip install PythonAPI/dist/valdrada-0.1.0-py3-none-any.whl numpy
python PythonAPI/examples/tutorials/first_steps.py

From a source checkout the scripts run without installing the package, since they add PythonAPI/ to the import path themselves. Every script takes --host and --port for a server on another machine or port, and --out for its output folder (default _out/<script name> in the current directory). The scripts put the world in synchronous mode, destroy the actors they spawn and restore the settings they changed before they exit.

Conventions

The world frame is ENU metres from 40.7831 N, 73.9712 W: x east, y north, z up. Rotations are pitch, yaw and roll in degrees, with yaw counter-clockwise from east, and a camera looks along its +x axis. The Manhattan street grid is rotated about 29° clockwise from true north, so its avenues run uptown at a yaw of about 61°; the scripts call this heading UPTOWN. The offset of an attached actor is in its parent's frame: x forward, y left, z up.