AnimalClue logo AnimalClue

AnimalClue: Recognizing Animals by their Traces

ICCV 2025 Highlight


Risa Shinoda1,2,4, Nakamasa Inoue3,4, Iro Laina5, Christian Rupprecht5, Hirokatsu Kataoka4,5

1The University of Osaka 2Kyoto University 3Tokyo Institute of Technology

4National Institute of Advanced Industrial Science and Technology (AIST) 5Visual Geometry Group, University of Oxford

arXiv 🤗 Dataset Code 🤗 Demo Poster
AnimalClue dataset: annotated footprints, feces, eggs, bones, and feathers of animals such as bobcat, American black bear, and whimbrel, with a taxonomy of 968 species

Example images of AnimalClue. We present AnimalClue, a dataset designed for identifying animal species based on their traces. Our dataset includes footprints, feces, eggs, bones, and feathers, totaling 159,605 bounding boxes from 968 animal species. We also annotate 22 traits such as habitat, diet, and activity pattern. We establish four benchmarks for evaluating models: classification, detection, instance segmentation, and traits classification.

🔔News

🔥[2025-08-25] We release the Demo on Hugging Face 🤗
🔥[2025-07-29] Introducing AnimalClue on arXiv! 🚀

Introduction

Wildlife observation plays an important role in biodiversity conservation, necessitating robust methodologies for monitoring wildlife populations and interspecies interactions. Recent advances in computer vision have significantly contributed to automating fundamental wildlife observation tasks, such as animal detection and species identification. However, accurately identifying species from indirect evidence like footprints and feces remains relatively underexplored, despite its importance in contributing to wildlife monitoring. To bridge this gap, we introduce AnimalClue, the first large-scale dataset for species identification from images of indirect evidence. Our dataset consists of 159,605 bounding boxes encompassing five categories of indirect clues: footprints, feces, eggs, bones, and feathers. It covers 968 species, 200 families, and 65 orders. Each image is annotated with species-level labels, bounding boxes or segmentation masks, and fine-grained trait information, including activity patterns and habitat preferences. Unlike existing datasets primarily focused on direct visual features (e.g., animal appearances), AnimalClue presents unique challenges for classification, detection, and instance segmentation tasks due to the need for recognizing more detailed and subtle visual features. In our experiments, we extensively evaluate representative vision models and identify key challenges in animal identification from their traces.

Video

AnimalClue

Overview

Annotation examples from the AnimalClue dataset: bounding boxes on footprints and segmentation masks on feces, eggs, bones, and feathers, labeled by species

Our dataset consists of 968 species, 200 families, and 65 orders. It includes a total of 159,605 bounding boxes across five trace types:

  • Footprints: 18,291 boxes from 7,581 images, covering 117 species, 46 families, and 20 orders
  • Feces: 18,932 boxes from 6,433 images, covering 101 species, 46 families, and 21 orders
  • Bones: 16,553 boxes from 12,908 images, covering 269 species, 112 families, and 45 orders
  • Eggs: 29,434 boxes from 9,394 images, covering 283 species, 67 families, and 20 orders
  • Feathers: 76,395 boxes from 60,491 images, covering 555 species, 89 families, and 30 orders

The total number of bounding boxes matches that of the classification dataset, while the number of images aligns with the detection and segmentation datasets.

Comparisons with Existing Benchmarks

While direct animal identification has been extensively studied, there remains significant potential to explore indirect methods—such as identifying animals through the traces they leave behind. Our AnimalClue dataset comprises five trace types and 968 species, with a total of 159,605 bounding boxes. It supports a wide range of tasks and includes fine-grained annotations.

Table comparing AnimalClue with previous animal track datasets by trace types, number of species, bounding boxes, and supported tasks

Comparison with previous animal tracking datasets. CLS, DET, and SEG indicate classification, detection, and instance segmentation, respectively. Our AnimalClue contains diverse species and more number of bounding boxes.

Statistics

Bar chart of annotation counts per species in AnimalClue, stacked by footprint, feces, bone, egg, and feather

Species distribution in our AnimalClue. We illustrate the top 100 most frequent species, with the vertical axis representing log frequency.

Experiment Results

Classification

Table of species, family, and order classification accuracy on AnimalClue for all, frequent, and rare categories

Classification accuracy for all, frequent, and rare categories of animal species. Throughout the species, family, and order categorization, Swin-B model tends to be higher accuracies on AnimalClue.

Visualization of t-SNE

t-SNE plots of animal trace image features from CLIP, BioCLIP, and fine-tuned CLIP on AnimalClue

Visualization of t-SNE. By using a labeled dataset specialized for observing indirect animal clues, the separability among categories have been improved. When visualized in the feature space, the categories are better distinct.

Detection

Table of object detection results (mAP@50-95) on AnimalClue footprint, feces, egg, bone, and feather data

Detection results on footprints, feces, eggs, bones, and feathers datasets (mAP@50-95).

Segmentation

Table of instance segmentation results (mask mAP@50-95) on AnimalClue feces, egg, bone, and feather data

Segmentation results on feces, eggs, bones, and feathers datasets (Mask mAP@50-95).

Traits Classification

Table of trait classification accuracy and F1 on AnimalClue, including diet type, activity pattern, habitat, herding, and predator

Traits classification results on footprints, feces, eggs, bones, and feathers datasets (Acc/F1 score).

Visualization of detection results

YOLOv11 detection examples on animal footprints, feces, eggs, bones, and feathers with predicted species such as common raccoon and wild turkey

Visualization of YOLOv11 detection results. The green bounding box denotes the correct detection, and the red bounding box denotes the wrong detection

Visualization of segmentation result

YOLOv11 instance segmentation examples on animal feces, eggs, bones, and feathers with predicted species such as gray wolf and red-winged blackbird

Visualization of YOLOv11 instance segmentation results. The green mask denotes the correct mask, and the red mask denotes the wrong mask.

BibTeX


@inproceedings{shinoda2025animalclue,
  title={AnimalClue: Recognizing Animals by Their Traces},
  author={Shinoda, Risa and Inoue, Nakamasa and Laina, Iro and Rupprecht, Christian and Kataoka, Hirokatsu},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={14776--14786},
  year={2025}
}