Project proposal details

Please look carefully through the proposal details below. If you are interested in the project then contact the supervisor, explaining why you are interested and any background which makes you a good fit for the project.

If this is an external project, the lead supervisor may have suggested someone at Imperial College or the NHM who could act as an internal supervisor and you should also contact them. If the project is external and no internal has been proposed then you must find an internal supervisor before starting the project.

Please pay close attention to any extra notes on requirements (such as being able to drive or to speak particular languages) or the application process. There may be specific limitations on the project availability: if there are then they will be clearly shown further down the page.

Project title
Using AI to annotate images from the deep-sea
Contact name
Chris Yesson
Project based at
Zoological Society of London
Project description
1. Using AI to annotate seabed images from Greenland

Benthic (seabed) habitats are an important part of the marine ecosystem. However, studying these environments can be challenging, particularly in deep-seas and polar environments, which means these are some of the least well understood habtiats on earth. Researchers at IoZ have been performing annual imaging surveys of the seabed in West Greenland since 2011 (https://www.zsl.org/greenland). We have been using seabed videos to describe, document and analyse life on the seabed (https://youtu.be/j84wFEJuseI) and assess the impact of trawling on benthic habitats (https://dx.doi.org/10.1093/icesjms/fsw206).

A fundamental part of this work is examining animals on the seabed seen in benthic imaging surveys. Often this requires looking at and annotating hundreds of images to record species observations. This is time consuming and quality of annotations can be dependent on the person assessing the images (https://doi.org/10.1016/j.pocean.2016.07.005). Artificial Intelligence and machine learning have been used in a few cases to identify and record animals on the seabed (https://doi.org/10.1371/journal.pone.0207498). This project will assess the abililty of the MAIA software to identify widespread fauna on the seabed in West Greenland, and use these data for community analysis to define, map and document the major biotic habitats of the region.

This is an opportunity to gain an understanding of seabed habitats by examining video footage from surveys of the Greenland continental shelf, with training in the image annotation plaform Biigle (https://doi.org/10.3389/fmars.2017.00083 - https://biigle.de), and learn about cutting edge technology to automatically identify fauna from images.
Date uploaded
2019-11-14

Project proposal limitations

The project proposer has indicated that there are some limitations to the availability of this project. It may only be available at certain times of year or suit a specific project length. It may also need skills taught to students on a particular course or courses.

Research project proposals are usually part of an active research programme. If supervisors have stated limitations to a proposal, then they are unlikely to have any flexibility. If you are very interested in the topic but have problems with the stated limitations, the supervisor may still be happy to talk to you about other options around the proposal, but you should not expect that any alternative arrangements can be made.

Project length limitations
3.5 months, 5 months
Available date limitations
Winter (January), Spring (April-May)