Skip to contents

A deep learning model was trained on the signed distance field of 3d bird beak scans. This dataset contain the learned latent codes that produce the bird beak shapes when passed through the trained companion neural network. The trained neural network is available from load_model() #d scans used to train the model were retrieved from the MarkMyBird project dataset (https://www.markmybird.org/).

Usage

bird_beak_codes

Format

bird_beak_codes

A 'pf' data frame (subclasses tibble) with 4,040 rows and 80 columns:

label

Node labels including species name for the tip labels

is_tip

Logical specifying whether the row represents a tip on the phylogeny

phlo

The phylogenetic flow column which stores the phylogenetic information

Common_name

The English common name for the bird species

Scientific

The scientific name for the bird species

Clade

Various traits of the bird species, see Source section to get more detailed information

BLFamilyLatin

Taxonomic family latin name

BLFamilyEnglish

Taxonomic family English common name

Order

Taxonomic order

OscSubOsc

Oscine or Sub-Oscine

X and Y

Two dimensional UMAP dimension reduction of the 64 latent variables

X0, Y0, and Z0

Three dimensional UMAP dimension reduction of the 64 latent variables

latent_code_1:latent_code_63

64 latent codes representing bird bealk shapes, estimated using a autodecoder neural network architecture