Scholarlink

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% 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% points; the mean relative error drops by 35%. We make the code and pre-trained models publicly available, enabling immediate use in ecology and conservation ( https://github.com/WildlifeDatasets/wildlife-tools ).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

WildFusion: Individual Animal Identification with Calibrated Similarity Fusion

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vojtěch Cermak,
  • Lukas Picek,
  • Lukáš Adam,
  • Lukáš Neumann,
  • Jiří Matas

摘要

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% 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% points; the mean relative error drops by 35%. We make the code and pre-trained models publicly available, enabling immediate use in ecology and conservation ( https://github.com/WildlifeDatasets/wildlife-tools ).