WildFusion: Individual Animal Identification with Calibrated Similarity Fusion

Vojtěch Cermak, Lukas Picek, Lukáš Adam, Lukáš Neumann, Jiří Matas
8/23/2024
cs.CV

No official code found. Best implementation: WildlifeDatasets/wildlife-tools (78 stars, MIT, last commit 2026-04-24).

Abstract

We propose a new method - WildFusion - for individual identification of a broad range of animal species. The method fuses deep scores (e.g., MegaDescriptor or DINOv2) and local matching similarity (e.g., LoFTR and LightGlue) to identify individual animals. The global and local information fusion is facilitated by similarity score calibration. In a zero-shot setting, relying on local similarity score only, WildFusion achieved mean accuracy, measured on 17 datasets, of 76.2%. This is better than the state-of-the-art model, MegaDescriptor-L, whose training set included 15 of the 17 datasets. If a dataset-specific calibration is applied, mean accuracy increases by 2.3% percentage points. WildFusion, with both local and global similarity scores, outperforms the state-of-the-art significantly - mean accuracy reached 84.0%, an increase of 8.5 percentage points; the mean relative error drops by 35%. We make the code and pre-trained models publicly available5, enabling immediate use in ecology and conservation.

View on arXivView PDF

Code Implementations(1)

WildlifeDatasets/wildlife-toolsCommunity100%
MIT

An open-source toolkit for animal re-identification

7820Python, Jupyter NotebookApr 17, 20236 months agoMIT

Similar Papers

Multispecies Animal Re-ID Using a Large Community-Curated Dataset

87% match1 repo252024

Understanding the Impact of Training Set Size on Animal Re-identification

85% match2024

OpenAnimals: Revisiting Person Re-Identification for Animals Towards Better Generalization

84% match2024

Categorical Keypoint Positional Embedding for Robust Animal Re-Identification

84% match2024

Hierarchical Re-Classification: Combining Animal Classification Models with Vision Transformers

83% match2025

Transfer Learning for Wildlife Classification: Evaluating YOLOv8 against DenseNet, ResNet, and VGGNet on a Custom Dataset

83% match2024

Cite this paper

@article{cermak2024wildfusion,
  title  = {WildFusion: Individual Animal Identification with Calibrated Similarity Fusion},
  author = {Vojtěch Cermak and Lukas Picek and Lukáš Adam and Lukáš Neumann and Jiří Matas},
  year   = {2024},
  eprint = {2408.12934},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2408.12934v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2408.12934.svg)](https://codeofpaper.com/paper/2408.12934)

Discussion