Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Siamese Networks for Cat Re-Identification: Exploring Neural Models for Cat Instance Recognition

Tobias Trein, Luan Fonseca Garcia

TL;DR

This work tackles automated re-identification of urban street cats to support welfare initiatives like Hello Street Cat. It evaluates Siamese-network architectures with VGG16, EfficientNet, and MobileNet backbones under contrastive and triplet losses on a 2,796-image, 69-cat HelloStreetCat dataset, considering top and front viewpoints. The key finding is that VGG16 with contrastive loss on top-view images delivers the strongest performance (up to 97% accuracy and high F1), aided by data augmentation; the approach yields a practical, scalable solution for community-driven welfare monitoring. The study also provides a new dataset and a modular, extensible framework that can be extended to real-time deployment and broader datasets to enhance cat population monitoring and welfare efforts.

Abstract

Street cats in urban areas often rely on human intervention for survival, leading to challenges in population control and welfare management. In April 2023, Hello Inc., a Chinese urban mobility company, launched the Hello Street Cat initiative to address these issues. The project deployed over 21,000 smart feeding stations across 14 cities in China, integrating livestreaming cameras and treat dispensers activated through user donations. It also promotes the Trap-Neuter-Return (TNR) method, supported by a community-driven platform, HelloStreetCatWiki, where volunteers catalog and identify cats. However, manual identification is inefficient and unsustainable, creating a need for automated solutions. This study explores Deep Learning-based models for re-identifying street cats in the Hello Street Cat initiative. A dataset of 2,796 images of 69 cats was used to train Siamese Networks with EfficientNetB0, MobileNet and VGG16 as base models, evaluated under contrastive and triplet loss functions. VGG16 paired with contrastive loss emerged as the most effective configuration, achieving up to 97% accuracy and an F1 score of 0.9344 during testing. The approach leverages image augmentation and dataset refinement to overcome challenges posed by limited data and diverse visual variations. These findings underscore the potential of automated cat re-identification to streamline population monitoring and welfare efforts. By reducing reliance on manual processes, the method offers a scalable and reliable solution for communitydriven initiatives. Future research will focus on expanding datasets and developing real-time implementations to enhance practicality in large-scale deployments.

Siamese Networks for Cat Re-Identification: Exploring Neural Models for Cat Instance Recognition

TL;DR

This work tackles automated re-identification of urban street cats to support welfare initiatives like Hello Street Cat. It evaluates Siamese-network architectures with VGG16, EfficientNet, and MobileNet backbones under contrastive and triplet losses on a 2,796-image, 69-cat HelloStreetCat dataset, considering top and front viewpoints. The key finding is that VGG16 with contrastive loss on top-view images delivers the strongest performance (up to 97% accuracy and high F1), aided by data augmentation; the approach yields a practical, scalable solution for community-driven welfare monitoring. The study also provides a new dataset and a modular, extensible framework that can be extended to real-time deployment and broader datasets to enhance cat population monitoring and welfare efforts.

Abstract

Street cats in urban areas often rely on human intervention for survival, leading to challenges in population control and welfare management. In April 2023, Hello Inc., a Chinese urban mobility company, launched the Hello Street Cat initiative to address these issues. The project deployed over 21,000 smart feeding stations across 14 cities in China, integrating livestreaming cameras and treat dispensers activated through user donations. It also promotes the Trap-Neuter-Return (TNR) method, supported by a community-driven platform, HelloStreetCatWiki, where volunteers catalog and identify cats. However, manual identification is inefficient and unsustainable, creating a need for automated solutions. This study explores Deep Learning-based models for re-identifying street cats in the Hello Street Cat initiative. A dataset of 2,796 images of 69 cats was used to train Siamese Networks with EfficientNetB0, MobileNet and VGG16 as base models, evaluated under contrastive and triplet loss functions. VGG16 paired with contrastive loss emerged as the most effective configuration, achieving up to 97% accuracy and an F1 score of 0.9344 during testing. The approach leverages image augmentation and dataset refinement to overcome challenges posed by limited data and diverse visual variations. These findings underscore the potential of automated cat re-identification to streamline population monitoring and welfare efforts. By reducing reliance on manual processes, the method offers a scalable and reliable solution for communitydriven initiatives. Future research will focus on expanding datasets and developing real-time implementations to enhance practicality in large-scale deployments.
Paper Structure (24 sections, 3 figures, 7 tables)

This paper contains 24 sections, 3 figures, 7 tables.

Table of Contents

  1. introduction
  2. Theorectical Background
  3. Siamese Networks and CNNs
  4. VGG16
  5. EfficientNet
  6. MobileNet
  7. Data Augmentation
  8. Transfer Learning
  9. Related Work
  10. Localized parts detection
  11. Face and head detection
  12. Cat re-identification
  13. Dataset
  14. Experiments
  15. Experimental Setup
  16. ...and 9 more sections

Figures (3)

  • Figure 1: Example images of Mr. Egg used in the dataset. The left image shows a top view, while the right image shows a front view, illustrating the two perspectives captured for each individual cat.
  • Figure 2: Graphical visualization of the environment
  • Figure 3: Simplified visualization of the Siamese Network model using Mr. Chess and Ms. Princess as example