Feedforward Few-shot Species Range Estimation

Christian Lange, Max Hamilton, Elijah Cole, Alexander Shepard, Samuel Heinrich, Angela Zhu, Subhransu Maji, Grant Van Horn, Oisin Mac Aodha

TL;DR: A new low-shot method for estimating the spatial range of species

Abstract:

Knowing where a particular species can or cannot be found on Earth is crucial for ecological research and conservation efforts.
By mapping the spatial ranges of all species, we would obtain deeper insights into how global biodiversity is affected by climate change and habitat loss. However, accurate range estimates are only available for a relatively small proportion of all known species. For the majority of the remaining species, we typically only have a small number of records denoting the spatial locations where they have previously been observed. We outline a new approach for few-shot species range estimation to address the challenge of accurately estimating the range of a species from limited data. During inference, our model takes a set of spatial locations as input, along with optional metadata such as text or an image, and outputs a species encoding that can be used to predict the range of a previously unseen species in a feedforward manner. We evaluate our approach on two challenging benchmarks, where we obtain state-of-the-art range estimation performance, in a fraction of the compute time, compared to recent alternative approaches.

Lay Summary:

There are many animal species that we do not know much about. We may have only seen them a few times in a few locations, and otherwise we may only have a photo of them or a brief description of where they are found or what they look like. Estimating what areas on Earth these species live in (the range of the species) is important if we want to protect them or find out more about them.

We introduce a machine learning model called FS-SINR to help predict the range of a species from a few locations, or from text, or images, or any combination of those. Our model can take in information for a species that it has never seen before and predict the range straight away, without having to take time learning from the information we have for that species.

We find that our model makes ranges that are closer to those made by experts, compared to other recent models, and it also makes them more quickly, as we don't have to spend time teaching our model about the species before it can make the range.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/Chris-lange/fs-sinr

Primary Area: Applications->Everything Else

Keywords: species distribution modeling, SDM, species range modelling, spatial implicit neural representation, SINR, low-shot learning, few-shot learning

Submission Number: 12159