Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Ohio Small Animals — Species Classroom Subset v1.0
Summary
A 16-species subset of the LILA Ohio Small Animals camera trap dataset for use in the CAS Deep Learning — Computer Vision course. The cameras look straight down into a bucket that snakes, small mammals, lizards and ground birds are guided into by drift fences. The train, validation and test images come from different sites, so the test score measures how well a model works at sites it has not seen.
Source
- Dataset: Ohio Small Animals (AHDriFT camera traps)
- URL: https://lila.science/datasets/ohio-small-animals/
- License: Community Data License Agreement — Permissive variant 1.0
- Attribution: Balasubramaniam S., "Optimized Classification in Camera Trap Images: An Approach with Smart Camera Traps, Machine Learning, and Human Inference", Master's thesis, The Ohio State University, 2024
License
This subset is published under the Community Data License Agreement — Permissive, Version 1.0, the same license as the source dataset. If you use it, please cite the source above.
Notice of changes: these files are modified from the original Ohio Small Animals
data. Images were selected, renamed, resized to max 1024 px and re-encoded as JPEG,
and the labels and metadata were rewritten into the metadata.csv files (see
Curation below).
Curation
- Selected classes: every species with at least 500 images at 10 or more sites.
Left out:
empty(only 5 sites, all with one camera setup, so the camera would give away the label) andinvertebrate(a catch-all of mostly tiny insects and spiders). - Split strategy: by site — all cameras of one deployment (e.g. KPC1/KPC2/KPC3, or the N/S pair of a 2022 grid point such as KILC4N/KILC4S) go to the same split. Sites were assigned so that each species has about 70/15/15 of its sequences in train/val/test, with at least a few sequences in val and test.
- Sampling: at most 800 train / 200 val / 200 test images per species. Images
are taken one per sequence first, then a second and third evenly spaced frame
from the same sequence if needed (at most 3 per sequence). Use
seq_idto group frames of one sequence. - Image resolution: resized from 2048 × 1440 to max 1024 px on the longest side (1024 × 720), JPEG quality 90
- Format: JPEG, ImageFolder layout (
<split>/<label>/<filename>.jpg)
Statistics
Images (sequences) per split:
| Class | Train | Val | Test |
|---|---|---|---|
| eastern_gartersnake | 800 (800) | 200 (200) | 200 (200) |
| song_sparrow | 800 (800) | 64 (23) | 200 (200) |
| meadow_vole | 800 (800) | 200 (104) | 200 (200) |
| white-footed_mouse | 800 (800) | 200 (200) | 200 (200) |
| northern_house_wren | 800 (529) | 188 (70) | 104 (38) |
| common_five-lined_skink | 800 (545) | 200 (200) | 43 (15) |
| masked_shrew | 800 (800) | 200 (158) | 200 (200) |
| eastern_cottontail | 780 (311) | 60 (23) | 130 (58) |
| long-tailed_weasel | 395 (152) | 82 (31) | 112 (42) |
| woodland_jumping_mouse | 682 (300) | 188 (116) | 51 (24) |
| plains_gartersnake | 519 (181) | 117 (42) | 97 (36) |
| eastern_massasauga | 349 (120) | 64 (23) | 98 (33) |
| virginia_opossum | 133 (46) | 75 (25) | 43 (15) |
| common_yellowthroat | 145 (59) | 60 (21) | 49 (19) |
| n._short-tailed_shrew | 359 (149) | 52 (23) | 49 (20) |
| dekay's_brownsnake | 224 (79) | 50 (18) | 46 (16) |
| Total | 9186 | 2000 | 1822 |
Sites: 43 train, 11 val, 8 test.
Notes:
- The classes are imbalanced (133 to 800 training images).
- Most species occur at only 2–3 of the val/test sites, so per-class val/test scores rest on few sites and are noisy for the small classes.
- Many images are night-time infrared (grayscale).
- Labels are per sequence in the source, so some frames show the animal only partly.
Sidecar files
metadata.csv— one row per image withimage_id,file_name(relative to the dataset folder),split,label,site,location(camera),seq_id,frame_num,datetime,source_file,width,height<split>/metadata.csv— the same rows withfile_namerelative to the split folder; these are what Hugging Facedatasetsreads
Usage in notebooks
from huggingface_hub import hf_hub_download
import tarfile
archive = hf_hub_download(
"marco-willi/camera-trap-osu-small-animals",
"osu_small_animals.tar.gz",
repo_type="dataset",
)
with tarfile.open(archive) as tar:
tar.extractall(DATA_PATH)
With Hugging Face datasets (the val folder becomes the validation split):
from datasets import load_dataset
ds = load_dataset("imagefolder", data_dir=DATA_PATH / "osu_small_animals")
ds = ds.class_encode_column("label") # species name -> ClassLabel
With torchvision, ImageFolder(DATA_PATH / "osu_small_animals" / "train") works
directly; the metadata.csv file in each split folder is ignored.
- Downloads last month
- 46