CARE Toolkit
We introduce CARE (Clip-based Animal RE-identification), an advanced, AI-driven web toolkit designed to enhance wildlife conservation efforts by enabling precise, scalable animal re-identification. We hope to make this technology available to the wider community, empowering researchers, conservationists, and enthusiasts alike to identify and track animals effortlessly, advancing wildlife conservation and ecological research.
ANimal Re-identification
The platform allows users to upload their images, to which they can browse, view, and edit with ease. Uploaded images can be processed through the Detection Model to automatically identify animal species. Following this, the Re-Identification Model analyses the detected animals, distinguishing each unique individual across the selected images. All categorised animal images are organised in a results gallery which can be filtered, edited, deleted, and exported.
Classification
The initial step in re-identification involves detecting animals within raw images. CARE enables users to upload images to the platform, where our animal detection model processes each image to identify and classify animals, providing precise coordinates for each detection. The user-friendly Detection Gallery displays a clear overview of all identified animals, allowing users to easily view and confirm classifications for each image.
Re-identification
Once images are classified, users can proceed with re-identification. CARE organises each uniquely identified animal into its own folder, making it easy for users to manage their data. Within the platform, users can download, rename, or delete each folder as needed.
Video Demo
Technology
We used a lightweight backend implementation using Node.js through Express.js, which is also easily deployable to different systems. For the frontend, React was used, allowing for easy connection to the Node.js backend. Lastly, the model, housed on a AI server built with Flask, is interfaced with the backend server to process the images in parallel for re-identification. The implementation of this architecture allows the platform to be scalable, supporting high volumes of image uploads and large datasets, as well as quick processing speeds.
Case Studies
Stoat re-identification
Stoats pose a relentless threat to Waiheke Island’s native wildlife, their predatory prowess devastating populations of ground-nesting birds like the dotterel, ōi (grey-faced petrel), and kororā (little blue penguin), and hole-nesting birds such as kākā. These vulnerable species have limited defenses against stoat predation, particularly when rearing young. Recognising this urgent ecological crisis, the Te Korowai o Waiheke; Towards a Predator Free Waiheke program has prioritised the complete removal of stoats from the island.
To help with this endeavour, we utilised CARE to efficiently re-identify the few remaining stoats in the island. Our system tackles real-world challenges like image blur, nighttime photography, and partially obscured animals to accurately identify individual stoats. This critical data empowers conservationists with insights into stoat population dynamics, movement patterns, and the effectiveness of eradication strategies, ultimately safeguarding Waiheke’s cherished birdlife.
This project is carried out in collaboration with TAIAO, Pure Salt NZ, Te Korowai o’Waiheke, DOC, and Manaaki Whenua – Landcare Research.
Feral Cat Re-identification
Feral cats pose a severe threat to New Zealands’s native wildlife, preying on birds, bats, lizards, and insects, and have contributed to the decline of numerous endangered species. Their impact is particularly severe on endangered birds like the black stilt and kea, and they’ve driven multiple bird species to local extinction, especially on offshore islands. Additionally, feral cats can transmit toxoplasmosis, affecting native dolphins. The Department of Conservation (DOC) works to control feral cat populations using humane methods across conservation sites, but challenges remain due to reinvasion.
We will be collaborating with DOC and Manaaki Whenua – Landcare Research to fine-tune our CARE platform to re-identify feral cats across a particular region. We are currently in the process of finalising this new model, while it shows promising results.
Kiwi bird re-identification
Besides eradication of invasive species, various sectors can also utilise CARE in carefully monitoring endangered species out in the wild. In this front, we collaborated with Department of Conservation (DOC) and Manaaki Whenua – Landcare Research in re-identifying kiwi birds. This technology can streamline conservation efforts such as tracking and population monitoring that would otherwise be time consuming and cumbersome.
This is currently a work in progress.
Our researchers include Yihao Wu, Di Zhao, Yuzhuo Li, and Fiona Bautista, with the help of supervisors Prof Yun Sing Koh, Prof Gillian Dobbie, Dr Daniel Wilson, Assoc Prof Al Glen, Dr Jingfeng Zhang, and Dr Brent Martin. We would like to acknowledge our partners Te Korowai o Waiheke, Department of Conservation, and Manaaki Whenua – Landcare Research.