pycolmap_constraints

Overview

pycolmap_constraints is a Python script that leverages pycolmap and pyceres to optimize 3D reconstructions using pose priors, rigs, and bundle adjustment.

Usage

Running the Script

To run the main.py script, use the following command:

python main.py --config /path/to/config.ini

Configuration File (config.ini)

The script requires a configuration file in .ini format to specify input, output, and optimization parameters. Below is a description of the sections and parameters:

[General]

• input: Path to the input folder containing the reconstruction data.
• output: Path to the output folder where the optimized reconstruction will be saved.
• priors_delimiter: Delimiter used in the priors file (e.g., ;).

[Problem]

• bundle: Boolean (true or false) to enable or disable bundle adjustment.
• priors: Path to the file containing pose priors.
• priors_covariance: Covariance matrix for the priors, specified as a list (e.g., [0.1, 0.1, 0.1, 100, 100, 100]).
• use_priors_rotations: Boolean to indicate whether to use rotation priors.
• priors_rotations_from_reconstruction: Boolean to use rotations from the reconstruction instead of the priors file.
• transform_reconstruction: Boolean to apply an affine transformation to align the reconstruction with the priors.
• fixed_cameras_ids: List of camera IDs to keep fixed during optimization (e.g., [1, 2, 3]).
• fix_3D_points: Boolean to fix 3D points during optimization.
• rigs_file: Path to the file containing rig configuration.

[Solver]

• linear_solver_type: Type of linear solver to use (e.g., SPARSE_SCHUR, DENSE_QR).
• minimizer_progress_to_stdout: Boolean to print solver progress to the console.
• num_threads: Number of threads to use for optimization.

Example Configuration File

Below is an example config.ini file:

[General]
input = /path/to/input
output = /path/to/output
priors_delimiter = ;

[Problem]
bundle = true
priors = /path/to/priors.txt
priors_covariance = [0.1, 0.1, 0.1, 100, 100, 100]
use_priors_rotations = true
priors_rotations_from_reconstruction = false
transform_reconstruction = true
fixed_cameras_ids = [1, 2, 3]
fix_3D_points = false
rigs_file = /path/to/rigs.json

[Solver]
linear_solver_type = SPARSE_SCHUR
minimizer_progress_to_stdout = true
num_threads = 4

Output

The optimized reconstruction will be saved in the specified output folder. Additional files such as optimized.txt, cloud_optimized.txt, and transformation_matrix.txt will contain the optimized camera positions, 3D points, and transformation matrices, respectively.

Dependencies

• pycolmap
• pyceres
• numpy
• configparser
• argparse