Datasets:

---

imageomics

/

ToL-EDA

like 0

Follow

HDR Imageomics Institute 156

While looking for a hybrid subset for the HDR Anomaly Data Challenge, I found that we had some plants mislabeled (from family up) to the phyla Artrhopoda and Chordata (possibly others).

From the discussion on GitHub:

It seems at least two of the genera that were potentially mislabeled on the higher order are known homonyms indicated in the IRMNG list (see genus level L and genus level V):

Laelia
Laelia J. Lindley, 1831 · Plantae (Kingdom) > Tracheophyta (Phylum) > Magnoliopsida (Class) > Asparagales (Order) > Orchidaceae (Family)
X Laelia Adanson, 1763 accepted as Laelia Adanson, 1763 · Plantae (Kingdom) > Tracheophyta (Phylum) > Magnoliopsida (Class) > Brassicales (Order) > Brassicaceae (Family)
X Laelia Persoon, 1806 accepted as · Plantae (Kingdom) > Tracheophyta (Phylum) > Magnoliopsida (Class) > Brassicales (Order) > Brassicaceae (Family)
Laelia Stephens, 1828 · Animalia (Kingdom) > Arthropoda (Phylum) > Insecta (Class) > Lepidoptera (Order) > Erebidae (Family)

and

Viola
Viola Evans, 1953 · Animalia (Kingdom) > Arthropoda (Phylum) > Insecta (Class) > Lepidoptera (Order) > Hesperiidae (Family)
Viola Linnaeus, 1753 · Plantae (Kingdom) > Tracheophyta (Phylum) > Magnoliopsida (Class) > Malpighiales (Order) > Violaceae (Family)

This may be part of the reason we had so many misclassifications into Arthropoda.

There is a notebook that partially explores this now (notebooks/make_ToL_hybrids_Arthropoda_EDA.ipynb) and two associated files (data/taxa_files/ToL_hybrids_Arthropoda.csv and data/taxa_files/sample_arthropoda_mislabels.txt which is a list of filenames for the sample images in data/taxa_files/sample_images/).