dracor

Travis build status CRAN status GitHub Downloads

The goal of dracor is to allow decoding of the Draco compressed meshes in R. This is done by wrapping the Draco C++ decoding library with the assistance of the Rcpp package.

The original motivation for dracor was decoding neuroglancer meshes of neurons for example as used by https://flywire.ai/.

Installation

dracor is available from CRAN:

install.packages('dracor')

but you can also install the development version like so:

remotes::install_github("jefferis/dracor")

Example

This is a basic example using a sample from the draco repository

library(dracor)
# get sample file from draco repository
carurl='https://github.com/google/draco/blob/master/testdata/car.drc?raw=true'
car.m3d=dracor::draco_decode(carurl)
str(car.m3d)
#> List of 2
#>  $ vb: num [1:4, 1:1856] 1.54 1.65 -1.21 1 1.57 ...
#>  $ it: int [1:3, 1:1744] 1 2 3 3 2 4 4 2 5 5 ...
#>  - attr(*, "class")= chr [1:2] "mesh3d" "shape3d"

rgl is the most widely used R package for 3D visualisation. By default we return meshes as rgl mesh3d objects, which can then be displayed by rgl or manipulated by a range of R packages including Rvcg.

# install.packages("rgl")
# convert to rgl mesh3d format
rgl::shade3d(car.m3d, col='red')
# set a nice viewpoint
rgl::view3d(theta = 60, fov=0, zoom=.7)

You must enable Javascript to view this page properly.

Some details

dracor is deliberately intended as a minimal decoder package without any dependencies besides the Rcpp package. It accepts raw bytes, a file or a URL as input and can produce either an rgl mesh3d object as output or a list containing points and 0-indexed faces. It essentially replicates the most basic decoding ability of the draco_decoder command line tool.

If you just want a result as close as possible to what the Draco library would give then set mesh3d=FALSE

car.m=dracor::draco_decode(carurl, mesh3d=FALSE)
str(car.m)
#> List of 2
#>  $ points: num [1:3, 1:1856] 1.54 1.65 -1.21 1.57 1.77 ...
#>  $ faces : int [1:3, 1:1744] 0 1 2 2 1 3 3 1 4 4 ...

Acknowledgements

Many thanks to the authors of: