This year’s MTRU Research Day was a resounding success, bringing together industry experts, academic researchers and PhD students to share insights into how marine technology and artificial intelligence are shaping the future of conservation and digital heritage. The day was packed with engaging talks, interactive demos and hands-on experiences.
The morning kicked off with Simon Brown’s presentation on photogrammetry of the Thistlegorm shipwreck offering a fascinating look at preserving marine heritage through 3D modelling and using forensic approaches to understand how the ship sank and has changed over time since she went down.
His team visited the wreck site a number of times, building up their understanding of the wreck site layout and where objects of interest were using accurate underwater mapping equipment. The explosion when the ship sank blasted numerous heavy objects out of the boat that now rest 100s of meters away from the main site. The distance and orientation tell us something about the location and power of the blast that took the ship down.
The accuracy of the 3D model Simon’s team have developed is incredibly accurate, so much so that the original ship’s plans can be overlaid over the 3D model. This allows the team to precisely determine which parts of the ships are damaged and missing, in the case above, the damage to an internal bulkhead (identified by the dashed red line) which should be straight (indicated by the black line in the drawing).
Perhaps the most important work on the site is logging all the artefacts to preserve the site. It was common for dive boats to tie on to the wreck which over time has pulled wreckage around the site, some of which could be explosive munitions. You can see how the anchors and chains on the bow of the ship have been moved (red and orange). Read more about this incredible project in his book Diving the Thistlegorm: The Ultimate Guide to a World War II Shipwreck.
The first short talk session, chaired by Dr. Louis Clift, began with Jess Rhodes uncovering the complexity of Norfolk’s chalk reefs using 3D modelling and comparing ecological monitoring recommendations across software packages. Emma Poliakova presented her work on the “Hector Detector” for seabed exploration, highlighting the challenges of AI distinguishing between similar morphologies. Dr. Jon Chamberlain discussed whether citizen scientists can create 3D models of marine environments, showcasing some promising results in collaboration with Queen’s University Belfast. Finally, Samia Mohinta from the University of Cambridge captivated the audience with her talk on creating digital brain maps with artificial intelligence, drawing connections between marine technology and broader scientific advancements.
Lunch fueled discussions around the posters and demos in the MTRU lab. Attendees had the chance to get hands-on with the new Fifish E-Go ROV, thanks to Mantsbrite Marine. These innovative tools sparked plenty of interest in the potential of remote technologies for underwater exploration.
The afternoon session, chaired by Jess Rhodes, kicked off with Dr. Lara Howe from the Manx Wildlife Trust discussing how they use innovative technology in their conservation work. She shared some incredible footage from their thermal imaging drone that’s used for seal population monitoring around the Isle of Man, demonstrating how non-intrusive methods can enhance conservation efforts.
The above image is a video still from the thermal image camera on the drone, showing the birth of a seal pup deep in cave of the Calf of Man. This incredible moment could not have been captured in any other way and made a big splash on the media channels. This type of engaging footage complements the “hard science” of counting seal populations to monitor change.
Manx Wildlife Trust use other types of monitoring technology that biologists will be more familiar with. Drop down cameras and baited remote underwater video setups can be used to collect data on marine life without sending divers into the water. The footage collected can then be post-processed to automatically count and measure, for example, sea grass within the quadrat.
The shark tagging project in Manx waters has also been a great success, with individuals being tracked from the Isle of Man to other geographical areas. Their research has indicated that Manx waters may be used as a nursery ground.
The session continued with more presentations from PhD students. Grace Long presented her work on deep learning for shark detection in the deep sea. Richard Lay-Flurrie discussed the simulation of maritime environments to prepare operators for high-pressure conditions. Russ Connelly’s presentation on UK seahorses explored how AI can help detect these elusive creatures, emphasising why visual sightings alone are insufficient.
In the final session, Steven Dey of ThinkSee3D presented some of his recent work in marine-related spatial heritage projects, showcasing cutting-edge methods in high resolution photogrammetry and hyper-realistic 3D modelling.
To kick us off he presented the background behind one of the most important papers recently published in measuring structural complexity for coral reef conservation (Young et al., 2017). Colleagues from Oxford University contacted him to discuss the idea of creating a digital twin of coral reefs and performing detailed computational measurements of the structure. With Steve’s background in physics and 3D modelling they collaborated to produce a low cost method of accurately measuring the complexity of coral reefs.
The work on coral reefs presents a method for capturing data using low cost equipment and this sparked a lively discussion about the “art of the possible”. Steve’s ambition to create the most realistic looking models has lead to some extraordinary results. Collaborative work with the University of Essex to recreate buildings lost to Italian earthquakes are so real you would find it hard to believe they weren’t a photo (see above!).
Not content with hyper-realistic photos, Steve took us for a tour around a lighthouse he scanned from the Falklands Islands. The model was rendered in Unreal Engine (the same technology Richard presented on earlier), allowing the user to explore the environment in high detail, whilst being lashed with rain and wind.
Following his inspiring talk, Steve led an open discussion in the fishbowl session on future trends in marine technology. Emerging themes included:
- Differences between photogrammetry software: Metashape, Zephyr, Meshroom, Pix4D
- How the quality of input images impact the final output: what can citizen scientists achieve? what should scientists aspire to to extract data? what is the highest level we can achieve?
- Video vs still images for photogrammetry – the debate rages on!
- The importance of lighting and colour balance, and whether it can be corrected in post-processing
- Machine learning to predict multiple variables beyond semantic segmentation
- Visualising segmentation in 3D
- Scaling up approaches
- The need for humans in the loop to verify AI identification
- Aligning traditional observation with AI observations to prevent shifting baselines
- The possibilities of virtual environments for outreach, training and education
In the evening we joined Ipswich Scuba for a pool session, who offered a try dive for first-time SCUBA divers Emma and Richard. At the other end of the pool Jon, Russ, and Louis trialed the ROVs in action, demonstrating their versatility and potential for marine research.
The MTRU Research Day highlighted the exciting intersections of technology, AI and marine conservation. We are inspired to explore new frontiers in preserving and understanding our oceans and innovate in this area. We’re already looking forward to the next year’s event!