Hugging Face's logo

Datasets:
davanstrien
/
ena24-detection

Tasks:
Object Detection
Modalities:
Image
Formats:
parquet
Size:
1K - 10K
Tags:
wildlife
camera-trap
conservation
object-detection
bounding-boxes
animals
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
ena24-detection / README.md
davanstrien's picture
davanstrien HF Staff
Upload README.md with huggingface_hub
6420f9a verified
preview code
|
raw
history blame contribute delete
3.18 kB
metadata 
license: cdla-permissive-1.0
task_categories:
  - object-detection
tags:
  - wildlife
  - camera-trap
  - conservation
  - object-detection
  - bounding-boxes
  - animals
  - ecology
size_categories:
  - 1K<n<10K

ENA24 - Eastern North America Camera Trap Object Detection

Camera trap images with bounding box annotations for 23 wildlife species from Eastern North America.

Dataset Description

This dataset contains 8,789 camera trap images with 9,772 bounding box annotations across 23 species. Images are 1920x1080 resolution. Annotations are in COCO format (xywh bounding boxes).

Originally collected by the University of Missouri and distributed via LILA BC (Labeled Information Library of Alexandria: Biology and Conservation).

Species (23 categories)

Species Count Species Count
American Crow 1,278 Eastern Cottontail 340
American Black Bear 959 Bobcat 333
Chicken 732 Eastern Gray Squirrel 319
Dog 726 Eastern Chipmunk 311
Virginia Opossum 725 Striped Skunk 297
Domestic Cat 482 Northern Raccoon 292
Grey Fox 438 Bird 225
Wild Turkey 427 Woodchuck 206
Red Fox 413 Vehicle 138
White-Tailed Deer 381 Horse 63
Coyote 344 Human 0
Eastern Fox Squirrel 343

Dataset Structure

Each example contains:

  • image: Camera trap photograph (PIL Image)
  • image_id: Unique image identifier
  • width / height: Image dimensions
  • objects: Object detection annotations
    • bbox: List of bounding boxes in COCO format [x, y, width, height]
    • category: List of species labels (ClassLabel)
    • area: List of bounding box areas

Usage 

from datasets import load_dataset

dataset = load_dataset("davanstrien/ena24-detection")

Train an object detector

This dataset is compatible with the Hugging Face vision-trainer skill for fine-tuning object detection models:

hf jobs uv run --flavor l4x1 \
  https://huggingface.co/datasets/hugging-face-vision-trainer/scripts/raw/main/object_detection_training.py \
  --dataset_name davanstrien/ena24-detection \
  --model_name_or_path facebook/detr-resnet-50 \
  --output_dir wildlife-detector \
  --do_train --do_eval \
  --push_to_hub

Source and License 

@misc{ena24,
  title={ENA24 Detection Dataset},
  author={University of Missouri},
  year={2016},
  url={https://lila.science/datasets/ena24detection/}
}

Acknowledgments

This dataset was originally hosted on LILA BC, a repository for labeled information supporting conservation biology research. Redistributed here under the CDLA Permissive 1.0 license.