Researchers at MIT have developed a novel approach to documenting ocean biodiversity by combining artificial intelligence with underwater photography.
The initiative—called LOBSTgER, short for Learning Oceanic Bioecological Systems Through Generative Representations—aims to capture and visualize marine life in the rapidly changing Gulf of Maine. By training custom generative models on curated, high-resolution underwater images, the project offers a new way to tell ecological stories with both scientific accuracy and artistic depth.
Original images photographed by Keith Ellenbogen to document New England's marine life and serve as the training foundation for LOBSTgER's generative models. Image Credit: Keith Ellenbogen
A Changing Gulf of Maine
The Gulf of Maine, one of the most biodiverse marine regions on the planet, is warming faster than 99 % of the global ocean—a trend that poses a serious threat to its ecosystems. LOBSTgER, co-led by underwater photographer Keith Ellenbogen and engineer Andreas Mentzelopoulos, is designed to confront this challenge through a unique blend of art, science, and technology.
The team trains AI models exclusively on Ellenbogen’s field-based imagery, ensuring the outputs remain both biologically accurate and visually compelling.
Drawing on marine biology, computational AI, and visual arts, the project reflects MIT’s interdisciplinary ethos. LOBSTgER also addresses practical limitations in underwater photography, such as poor visibility, lighting distortion, and the unpredictability of marine encounters, while enabling immersive representations of ocean life that would otherwise be difficult to capture.
Where Art Meets AI
At the heart of the project is a commitment to preserving the authenticity of marine environments while expanding how they can be visualized. Ellenbogen’s photographs, featuring species like lion’s mane jellyfish and ocean sunfish, form the training data for LOBSTgER’s generative models.
Each image is crafted with both aesthetic care and scientific precision, capturing species-specific details such as coloration, behavior, and natural lighting conditions. These images, part of Ellenbogen’s Space to Sea collection, are then used to teach the AI how to generate synthetic but ecologically faithful scenes.
Mentzelopoulos has developed custom diffusion models that not only replicate biodiversity but also emulate Ellenbogen’s artistic style. By learning from thousands of curated images, the AI absorbs fine visual cues like light refraction or suspended particulate matter to produce outputs that feel realistic and immersive.
The system operates in two main modes: one for generating entirely new underwater scenes from scratch and another for enhancing real photos by recovering obscured details. This dual capability allows LOBSTgER to simulate rare encounters or refine imperfect shots, expanding what’s possible in marine storytelling.
The team also stressed here that AI isn’t being used to replace traditional photography; instead, it is augmenting it, offering new ways to visualize complex ecosystems and deepen public understanding.
Innovation with Impact
What sets LOBSTgER apart technically is its foundation in original latent diffusion models, trained from scratch to avoid the biases and noise often found in general-purpose datasets. Building these models requires extensive computational resources, including hundreds of hours of tuning to align with the project’s creative and scientific objectives. As a result, the AI can, for example, transform a murky lobster image into a clear, detailed scene or create photorealistic renderings of marine species without direct reference input.
But the project’s impact extends well beyond technical innovation. By visualizing species that are endangered or sensitive to climate change—such as blue sharks and lobsters—LOBSTgER turns abstract ecological data into engaging visual narratives. With the Gulf of Maine warming at an unprecedented rate, the team hopes these images can help drive awareness and inspire conservation action.
Looking Forward
Looking ahead, the project is expected to scale its models to represent other marine ecosystems and collaborate with researchers to simulate environmental scenarios under climate stress. The team is also exploring open-source versions of its tools to support other photographers and conservationists, broadening access to AI-powered environmental storytelling.
LOBSTgER exemplifies how AI can help us see and understand ocean life in new ways. By merging scientific accuracy with artistic expression, the project offers a compelling approach to documenting ecosystems that are increasingly under threat. As the Gulf of Maine continues to warm, tools like this become more than just innovative; they become essential for communicating the urgency of environmental change.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.