FAQs

Yes, copyright of all images remains with the photographer. That’s why we ask you to identify the photographer as well as the submitter in each encounter submission.

We have committed to running the project until at least 2030.

The full user agreement is accessible in a number of ways and you should make yourself familiar with it. In short, by submitting photographs and seadragon sighting data you give SeadragonSearch permission to use this data for research and conservation purposes. You retain the copyright of your images but the intellectual property derived from the images (such as identifications) belongs to SeadragonSearch. Your personal details are never distributed beyond our system.

If we would like to share an image of yours on social media, we will always ask permission first, and credit you in the post. We make no assumptions that you would like to share this way, and we will always ask.

No. Many people participate through freediving, snorkeling and beach combing.

Any donations that come in through sponsorship are used for project management only. This includes things like web domain costs, website upkeep, contributor recognition and printing. It does not cover personal diving costs.

There are two ways we utilise this type of technology. The first way is to use a tool to automatically recognise a seadragon from its background. The second way is by comparing the patterns on each animal to other images in the database. We started the project using an algorithm called Hotspotter, which was then trained specifically on seadragon imagery, using a measure of visual texture. It created a score for a new query image, and then looked in the database for other images with similar scores, and provided these as a shortlist of possible matches.

In November 2024, we started using an algorithm called MIEW, which was trained simultaneously on many different species of animals (actually 64 of them, including seadragons). Like a really experienced graduate student that has worked on photo ID in many labs, it does a better job at predicting a match based on their vast experience. MIEW turns an annotation of a seadragon into a vector of numbers (2050 of them). Vectors of the same individual should be closer together in space (as a distance function) than vectors of other individuals. It also then provides a shortlist of possible matches. For both algorithms, a human from our team decides which of the suggested images might be a match, or if none are.