Exploring beneath the ocean's surface has always been a challenge, but a new wave of innovation is making the invisible visible. LOBSTgER, a cutting-edge project from MIT Sea Grant, is blending artificial intelligence with underwater photography to provide unprecedented insight into the Gulf of Maine. As climate change accelerates shifts in marine ecosystems, understanding and preserving ocean biodiversity is more urgent than ever.
Where Science, Art, and Technology Meet
At the heart of LOBSTgER, short for Learning Oceanic Bioecological Systems Through Generative Representations, is the use of generative AI to expand the limits of scientific storytelling. Drawing on the visually stunning work of photographer Keith Ellenbogen and led by PhD student Andreas Mentzelopoulos, the project trains custom AI models using a rich dataset of original underwater images. This ensures generated visuals are both artistically compelling and scientifically precise, rooted in real geographic locations and biological detail.
This fusion recalls how photography revolutionized natural documentation in the 19th century. Today, generative AI takes that legacy further, challenging our ideas of authenticity and offering fresh perspectives on marine life.
Image synthesis via reverse diffusion: This short video shows the de-noising trajectory from Gaussian latent noise to photorealistic output using LOBSTgER’s unconditional models. Iterative de-noising requires 1,000 forward passes through the trained neural network. Video: Keith Ellenbogen and Andreas Mentzelopoulos / MIT Sea Grant
Crafting an Accurate Visual Dataset
LOBSTgER’s AI models are grounded in Ellenbogen’s unique underwater photographs, which showcase New England’s marine diversity under difficult conditions. These images capture a spectrum of angles, lighting, and animal behaviors, resulting in a dataset rich in visual and biological nuance. To maintain integrity, Mentzelopoulos designed custom code that prevents external biases, ensuring the AI remains faithful to the documented environment.
Collaboration from Ocean to Algorithm
The project is a model of interdisciplinary teamwork. Ellenbogen ventures into unpredictable coastal waters to photograph creatures like blue sharks and lion’s mane jellyfish. Meanwhile, Mentzelopoulos translates these moments into AI models through rigorous training.
The result: a system that can generate new, photorealistic scenes and enhance existing images, whether by clarifying murky details or simulating encounters too rare or dangerous to photograph directly.
Left: Enhanced image of an American lobster using LOBSTgER’s image-to-image models. Right: Original image. Left: AI genertated image by Keith Ellenbogen, Andreas Mentzelopoulos, and LOBSTgER. Right: Keith Ellenbogen
Redefining Storytelling and Conservation
Rather than replacing traditional photography, LOBSTgER’s AI-generated visuals serve as a powerful complement. By infusing AI into the creative process, the project empowers researchers, artists, and educators to tell more comprehensive, emotionally engaging stories about the ocean. This approach streamlines image curation, bolsters public outreach, and provides new tools for underwater photographers tackling environmental and technical challenges.
Beyond their visual appeal, these images hold significant value for conservation and education. As ecosystems like the Gulf of Maine adapt to rapid change, compelling and accurate visuals become essential for raising awareness, inspiring stewardship, and fostering deeper public appreciation of marine life.
A Vision for the Future of Visual Science
The LOBSTgER team envisions a future where photography and generative AI exist on a seamless spectrum. While photography records authentic moments, AI extends those moments, unlocking new insights and imaginative scenarios rooted in scientific reality. This synergy offers a dynamic way to communicate the complexity and splendor of underwater worlds, motivating action in the face of environmental change.
Though still evolving, LOBSTgER’s blend of art, science, and technology demonstrates how merging direct observation with machine intelligence can unlock hidden ocean stories. As it matures, the project promises to push the boundaries of how we experience and understand life beneath the waves.
AI and Underwater Photography Are Illuminating Ocean Discovery