AI for fully automated identification of southern right whales to improve research and conservation outcomes

South Africa
Biology
$1,135
Pledged
12%
Funded
$10,000
Goal
31
Days Left
  • $1,135
    pledged
  • 12%
    funded
  • 31
    days left

About This Project

As a global consortium of SRW researchers, we rely on decades' worth of irreplaceable photo-identification datasets to assess conservation status. Unfortunately, current photo-ID processing systems rely on manual photo manipulation and matching. This leads to bottlenecks in photo processing and an inability to proceed with in-depth research projects. We therefore aim to develop an advanced fully automated AI-based photo-identification system specifically for SRWs.

Ask the Scientists

Join The Discussion

What is the context of this research?

Southern right whales were hunted from >100,000 individuals to a few hundred around 1920. Today, since their international protection from whaling, the species is recovering but remains at low levels. In some of the wintering areas, long-term photo-ID studies have been ongoing since the 1970s. While methodologies for these long-term studies vary, most involve annual aerial surveys that count and photo-identify SRWs in coastal waters based on overhead images of the callosity pattern on their head. These datasets have provided a wealth of information on behaviour, population abundance and trends over the past five decades. However, due to the large sizes, the currently existing semi-automated comparison processes are no longer viable, and an automated system is needed.

What is the significance of this project?

The ability to automatically identify southern right whales from photographs will have wide-ranging implications for research, conservation and whale-based tourism. From a research perspective, the automated identification of individuals will allow researchers to process ID images efficiently, work through bottlenecks of data, incorporate years of unprocessed data, and compile a global ID catalogue of the species. Such improved research outputs will have direct impacts on the conservation management of the species. Furthermore, the ability to instantly identify these whales from a tourism vessel using citizen science will allow for improved outreach and communication to tourists, increasing awareness of the wider community.

What are the goals of the project?

This project aims to develop a fully automated AI-based matching system for individual photo-identification of southern right whales. This will be a milestone in the SRW scientific community as this is a promising tool that has shown progress in the past decade for several whale and dolphin species, allowing us to steer away from non-user-friendly and labour-intensive photo-identification systems.

Budget

Please wait...

Financial resources are needed to support the work of AI developers to progress and trial an advanced, fully automated photo-identification system for southern right whales, ensuring efficient data processing for effective leveraging of existing long-term datasets. This will allow addressing priority research questions for species management. The work includes the development of a standardised annotated southern right whale photo-ID database using images from the three largest global datasets available (Argentina, South Africa and Australia, together > 150,000 images). Subsequently, this database will be used to develop AI algorithms and training models to ensure an automated photo-ID accuracy benchmark of 90% top-1 potential matches.

Partial funding has been made available through YaquPacha and the Nuremberg Zoo, to whom we are very grateful. However, more funds are needed to ensure completion of the project.

Endorsed by

Southern right whales have been studied for half a century by extremely dedicated researchers. Currently, they need novel software tools to learn more about the whales’ dynamics to help protect them in a rapidly changing world. The proposed software will be developed by highly qualified specialists and will have a broad impact on the work of numerous scientists across the southern hemisphere. Unique individual whale stories are waiting to be unfolded to help inform more efective conservation and managament strategies.

Project Timeline

Funding dependent, the full project is estimated to last 14 months, including 8 months of data annotation and creation of training datasets and 6 months of algorithm development and testing.

Oct 01, 2024

Project Launched

Jan 01, 2025

Data preparation: create database records for each identified SRW using image data from the three largest global datasets available (>10,000 individuals) 

Aug 31, 2025

Creating training dataset: Develop a standardised dataset with correctly oriented and annotated images ready for algorithm training.

Feb 01, 2026

Algorithm development: Develop algorithms for SRW recognition, train and test the models, provide and setup AI cloud server with GPU capabilities

Meet the Team

Southern Right Whale Consortium
Southern Right Whale Consortium

Team Bio

The Southern Right Whale Consortium (https://southernrightwhaleconsortium.org) will work in collaboration with AI developers in AT Design (https://data.marinemammals.gov.au/arwpic) to complete this project. All matching algorithms developed under this proposal will be made open source.

Southern Right Whale Consortium

The Southern Right Whale Consortium is an international collaboration with the ultimate goal to improve our understanding and conservation management of the southern right whale across the globe!

Lab Notes

Nothing posted yet.

Additional Information

The ability to identify individuals of wildlife populations allow researchers to investigate population parameters such as reproductive rates, survival chances, population growth or decline rates, movement patterns and population connectivity, to name but a few.


Project Backers

  • 5Backers
  • 12%Funded
  • $1,135Total Donations
  • $227.00Average Donation
Please wait...

See Your Scientific Impact

You can help a unique discovery by joining 5 other backers.
Fund This Project