Module: Gsplat::IO::Colmap

Defined in:
lib/gsplat/io/colmap.rb

Overview

COLMAP sparse reconstruction reader and camera conversion utilities.

Defined Under Namespace

Classes: Camera, Dataset, ImageRecord, Point3D

Constant Summary collapse

MODELS =

Supported COLMAP numeric camera model IDs.

{
  0 => "SIMPLE_PINHOLE",
  1 => "PINHOLE",
  2 => "SIMPLE_RADIAL",
  4 => "OPENCV"
}.freeze

Class Method Summary collapse

Class Method Details

.qvec_to_rotmat(qvec) ⇒ Object

Converts a COLMAP wxyz quaternion to a 3x3 world-to-camera rotation.



78
79
80
81
 
# File 'lib/gsplat/io/colmap.rb', line 78

def qvec_to_rotmat(qvec)
  quaternion = Numo::DFloat.cast(qvec).reshape(1, 4)
  Math::Quaternion.to_rotmat(quaternion)[0, true, true].dup
end

.read(path, data_factor: 1) ⇒ Object

Reads cameras, registered images, and points from a sparse model directory.



32
33
34
35
36
37
38
39
40
41
 
# File 'lib/gsplat/io/colmap.rb', line 32

def read(path, data_factor: 1)
  directory = model_directory(path)
  extension = model_extension(directory)
  Dataset.new(
    cameras: read_cameras(File.join(directory, "cameras.#{extension}"), data_factor: data_factor),
    images: read_images(File.join(directory, "images.#{extension}")),
    points3d: read_points3d(File.join(directory, "points3D.#{extension}")),
    path: directory
  )
end

.read_cameras(path, data_factor: 1) ⇒ Hash{Integer=>Camera}

Reads and scales camera calibration records from a bin/txt file.

Parameters:

  • path (String)
  • data_factor (Numeric) (defaults to: 1) —

    image downsampling factor

Returns:



48
49
50
51
 
# File 'lib/gsplat/io/colmap.rb', line 48

def read_cameras(path, data_factor: 1)
  validate_factor!(data_factor)
  raw_records(path, :cameras).transform_values { |record| camera_record(record, data_factor) }
end

.read_images(path) ⇒ Hash{Integer=>ImageRecord}

Reads registered image poses from a bin/txt file.

Parameters:

  • path (String)

Returns:



57
58
59
 
# File 'lib/gsplat/io/colmap.rb', line 57

def read_images(path)
  raw_records(path, :images).transform_values { |record| image_record(record) }
end

.read_points3d(path) ⇒ Hash{Integer=>Point3D}

Reads sparse points from a bin/txt file.

Parameters:

  • path (String)

Returns:



65
66
67
68
69
70
71
72
73
74
75
 
# File 'lib/gsplat/io/colmap.rb', line 65

def read_points3d(path)
  raw_records(path, :points3d).transform_values do |record|
    Point3D.new(
      id: record.fetch(:id),
      xyz: Numo::DFloat.cast(record.fetch(:xyz)),
      rgb: Numo::UInt8.cast(record.fetch(:rgb)),
      error: record.fetch(:error),
      track: record.fetch(:track).freeze
    )
  end
end