Crate city2ba

city2ba is a set of tools for generating synthetic bundle adjustment datasets.

Bundle Adjustment is a nonlinear global optimization used to reduce noise in structure from motion (SfM) and simultaneous localization and mapping applications (SLAM). See https://en.wikipedia.org/wiki/Bundle_adjustment for more details. Not many bundle adjustment datasets are freely available, so this package contains tools for generating synthetic ones. A bundle adjustment dataset contains a set of cameras, a set of 3D points, and a set of camera-point observations. The goal of a bundle adjuster is to minimizer the difference between the projection of each 3D point into each camera and the location where it was actually observed. This package provides ways for generating zero error (ground truth) datasets in the generate and synthetic modules, and ways to add noise to existing datasets (so they are no longer zero error) in the noise module.

This crate also provides command line tools for generating bundle adjustment datasets:

# Generate a problem from a 3D model
city2ba generate model.obj problem.bal --num-cameras 100 --num-points 200

# Add noise to the problem
city2ba noise problem.bal problem_noised.bal --drift-strength 0.001 --rotation-std 0.0001

# Generate a problem using a city block grid
city2ba synthetic problem.bal --blocks 4

# Convert a problem to a format for visualization
city2ba ply problem.bal problem.ply

Modules

generate	Functions to generate camera and point locations on a 3D model.
noise	Functions for adding noise to bundle adjustment problems.
synthetic	Functions for generating cameras and points without reference geometry.

Structs

BAProblem	Bundle adjustment problem composed of cameras, points, and observations of points by cameras.
SnavelyCamera	Camera expressed as Rx+t with intrinsics.

Enums

Error

Possible errors in generating and reading bundle adjustment problems.

Traits

Camera

A projective camera.