thompsonmj/test-fine-grained-challenges

Fine-Grained Challenges

Morphologically similar species image groups (including cryptic species) for evaluating and tuning fine-grained classifiers on top of BioCLIP ecosystem model embeddings.

The corpus lives in a Lance dataset and can be acted on as a whole, on any single morphologically similar species group independently, or on any single image. Images, embeddings (such as the 768-d BioCLIP 2 embeddings), taxonomic lineage, provenance, and the search indexes are provided in the same table:

data/challenges.lance/

Schema

Column	Type	Meaning
uuid	string	stable image id
challenge_group	string	which challenge group, e.g. Peromyscus
seen_in_training	list<string>	model-version ids whose training data included this image, e.g. ["bioclip-2", "bioclip-2.5"]; [] = held out from every listed model (see Considerations)
species	string	full binomial (Peromyscus maniculatus) if available, else null
kingdom…genus, scientific_name	string	taxonomy
common_name	string	common name if available, else null
source_dataset, source_id, publisher, basisOfRecord, img_type, resolution_status	string	record provenance
source_url	string	URL of the image (from TreeOfLife-200M provenance)
license_name, license_link, copyright_owner	string	per-image license + attribution (see Licensing)
emb_bioclip2	fixed_size_list<float32>[768]	BioCLIP 2 embedding (IVF_PQ-indexed)
image	binary	webp bytes (decode with PIL.Image.open(io.BytesIO(...)))

seen_in_training is per-image and is constant within a source cohort, so a single group can mix images seen in training with held-out images. See Considerations for using this data.

Groups

group	images	seen_in_training	notes
Peromyscus	1432	["bioclip-2", "bioclip-2.5"]	deermice and relatives; 1332 species-labeled + 100 genus-only

Usage

Install: uv pip install pylance polars pillow

import lance
import numpy as np
URI = "hf://datasets/imageomics/fine-grained-challenges/data/challenges.lance"
ds = lance.dataset(URI) # scans remotely over hf://, no full download

# Full dataset
ds.count_rows()

# One group independently (filter pushed down to only retrieve relevant rows and column projection to only retrieve relevant columns)
peromyscus = ds.scanner(
    filter="challenge_group = 'Peromyscus'",
    columns=["uuid", "species", "emb_bioclip2"],
).to_table()

# Select by training exposure (unseen test set vs. seen in training)
unseen = ds.scanner(filter="array_length(seen_in_training) = 0").to_table()
seen_by_bioclip2 = ds.scanner(
    filter="array_has(seen_in_training, 'bioclip-2')").to_table()

# Image bytes (raw webp), same pattern as the HF Lance image example
import io
from PIL import Image
row = ds.take([0], columns=["image", "species"]).to_pylist()[0]
img = Image.open(io.BytesIO(row["image"]))

# Materialize (i.e. download) a group locally for heavy/training use (avoids Hub rate limits)
lance.write_dataset(peromyscus, "./Peromyscus.lance")

Quick metadata sampling can also use HF datasets streaming (load_dataset("imageomics/fine-grained-challenges", streaming=True)); for embeddings, images, or vector search use the Lance API above.

Considerations for using this data

The seen_in_training field records the model versions whose training data included each image, so in-distribution evaluation can be separated from unseen tests:

group / cohort	seen_in_training	meaning
Peromyscus (GBIF/EOL via TreeOfLife-200M)	["bioclip-2", "bioclip-2.5"]	in the training corpus of both models
(held-out cohorts, e.g. camera-trap)	[]	novel to every listed model

The seen_in_training field is derived per image from its presence in TreeOfLife-200M, the training data for the listed models. It is a per-image membership fact for those model versions. For held-out evaluation, select array_length(seen_in_training) = 0, which corresponds to images novel to every listed model.

Models referenced: bioclip-2 (trained on an earlier TreeOfLife-200M snapshot) and bioclip-2.5 (trained on a later, expanded snapshot). The corresponding embedding columns use a filename-safe slug: value bioclip-2.5 corresponds to column emb_bioclip2p5.

Currently, only emb_bioclip2 is provided.

Licensing and attribution

The dataset compilation (curation, metadata, embeddings, and indexes) is released under CC0-1.0, matching imageomics/TreeOfLife-200M. This does not relicense the images.

Each image keeps its own license, recorded in its row (license_name, license_link, copyright_owner) from TreeOfLife-200M's provenance.parquet.

No-derivatives (*-nd) licenses are excluded at build time, since the stored webp and embeddings are derivative works. The remaining mix is mostly NonCommercial (*-nc-*), with some ShareAlike (*-sa-*), plain attribution, and public-domain images.

Use each image under its own license.

Per-license counts, computed from the data (URI as defined above):

import lance
import polars as pl
t = lance.dataset(URI).to_table(columns=["challenge_group", "license_name"])
pl.from_arrow(t).group_by(["challenge_group", "license_name"]).len()  # license counts, per group