Embedded Files
Hey there! Welcome to Scuba.Tech
Here you'll find our research, videos, course notes, notices, and local resources.
Our work is defined by three continuing goals, known as the Trident:
Increase community stewardship with our marine ecosystem — by bringing accessible oceanographic media content to viewers, and by running educational events to practice hands-on volunteering & field studies.
Promote fields of oceanography & marine biology — by mentoring future technologists and advocates through operational experience and classroom curriculum.
Research new digital workflows to combine historiography with technology — by investigating new methods to survey shipwrecks and marine biomes in 3D & 4D augmented reality, virtual reality, and finally by releasing products and techniques as public open-source.
Available Presentations and Workshops for 2026
PRESENTATIONS
Adventure & Archaeology: The Mazotos Expedition
- TOPIC: Photogrammetric workflows and impacts, expedition planning, working with ~400BC artifacts & amphorae, conservation and preservation, remote decompression operations — 1.5 hr
Tech in Tec: Diving the Stolt Dagali with an ROV
- TOPIC: ROV drone operations with dive teams on shipwrecks, ROV dive planning, ROV selection, ROV survey augmentation, accessibility and crewing, broadcast — 1 hr
Lush Seas: Diving with the Stellar Stellers of Vancouver
- TOPIC: Steller sea lions, PNW wildlife and interactions, travel and diving considerations — 45 min
WORKSHOP
ROV Operations Clinic
- In this workshop, we will take your class on an outdoor diving adventure, showcasing the latest in ROV construction, operations, imaging, and hybrid-diver operations. This program is completely custom to your class size and needs, so inquire for a bespoke plan. We will assist with all permitting and site planning, or we can join as a special guest with your existing event! Runtime ranges: ~ 3 hrs to 9 hrs
Introduction to Photogrammetric 3D Survey… Underwater!
- Ready to take your shipwreck diving and survey skills to the next level, or curious about art in rapid digital survey? Starting from the ground-up, we will review current survey techniques before introducing photogrammetric survey. With little more than a camera, an accurate baseline, the right software, and know-how, you can build scaled-3D models in just minutes after your dive! This technique has revolutionized land-based archaeology and is now being implemented underwater. All required software will be supplied, and we will go step-by-step through the entire workflow to create your first model. Handouts with notes will be provided, so you can focus on the task-at-hand…and have a convenient reference guide for the future! ~ 4 hrs
- Optional Extension: All of the above, plus 2 hours of equipment setup, a 2-hour pool session, and 2 hours of modeling the results. Students will leave with 2+ 3D models, fully-configured software, and the diving and photography skills to begin real survey projects. ~ 10 hrs
To book for presentations & workshops, suggest grants, contribute to open-source public research, or just ask a question: email hello{åt}scuba{døt}tech
For all solicitations and notices, email agent{åt}scuba{døt}tech
OUR LATEST UPDATES:
Scuba.Tech Open-Source Workflow Aligns 99.95% of Underwater Shipwreck Photos into a Single 3D Model
Danbury, CT and Jersey City, NJ — July 7, 2026
A year of volunteer, intern, and fellow research under projects 23Z OSIRIS and 23Z GSHARC cracks the hardest step of underwater NeRF and Gaussian splatting, with preliminary three-camera results reaching 100%
Scuba.Tech Incorporated ("Scuba.Tech"), a 501(c)(3) nonprofit research and education lab, today announced a milestone in underwater 3D imaging: an entirely open-source alignment workflow that stitched 99.95% of survey photographs of the Charleston Tug shipwreck (2,012 of 2,013 images from a single camera) into one coherent 3D reconstruction. The photos were captured in two to five feet of visibility with heavy marine growth and schooling fish, conditions that routinely defeat standard photogrammetry tools. A preliminary follow-up run across all three cameras of Scuba.Tech's Triscopic Camera Array registered 100% of the full 5,995-photo dataset in a single clean model of nearly 900,000 points, and is now undergoing final verification.
Charleston Tug fly-around render
Charleston Tug - Fly-Around RenderThe result comes from project 23Z, Scuba.Tech's twin open research tracks for photorealistic underwater reconstruction: OSIRIS (Open-Source Immersed Radiant Imaging Survey), which applies Neural Radiance Fields, and GSHARC (Gaussian Splatting for Heritage, Archaeology, Reefs, and Conservation), which applies Gaussian splatting to shipwrecks, artifacts, and reef habitats. Both techniques depend on the same fragile first step: Structure from Motion (SfM) alignment, in which software works out exactly where in space each photograph was taken. Underwater, turbidity, low light, and constantly moving marine life starve alignment algorithms of the visual anchors they need. Until now, difficult datasets forced the team out of its open pipeline and into external proprietary alignment software.
Over the past year, Scuba.Tech Alignment Researcher Jake Hiban, with assistance from Joshua Gross and Christopher Drew, treated the open-source COLMAP pipeline as a modular test bench. Its five sequential stages (finding distinctive keypoints in every photo, choosing which photos to compare, matching keypoints between image pairs, verifying those matches geometrically, and reconstructing camera positions and a sparse 3D point cloud) can each be swapped for newer components. The team benchmarked classical and modern learned methods, including SIFT, SuperPoint, ALIKED, DISK, and DeDoDe, against the Charleston Tug dataset, grading every run on how many photos landed in one connected model.
COLMAP's defaults fractured the full 3 camera dataset into seven disconnected partial models, with only 76.9% of images in the largest fragment: enough to build a model, but not enough for quality renders. The winning recipe, DeDoDe v2 feature extraction, NetVLAD image retrieval for pair selection, and classical nearest-neighbor ratio matching, produced one coherent model with virtually every photo included. The preliminary three-camera run paired the same recipe with the open-source GLOMAP reconstructor and ran for more than 25 hours on the team's donated hardware for a single experiment.
New workflow point cloud, Charleston Tug
COLMAP default point cloud, for comparison
"Every diver knows how variable the visibility can be down there, and how challenging that can be when you’re trying to share a shipwreck or reef with the world," said Christopher Drew, President of Scuba.Tech. "Teaching open software to see through it anyway is exactly the kind of problem our volunteers, interns, and fellows came here to solve. Everything we learn becomes open-source for the public good, so any school, museum, or research group can rebuild what lies beneath their own waters."
All experiments ran on Scuba.Tech's in-house, containerized Nerfstudio environment and personally donated gaming GPUs. Companion rendering studies on the Charleston Tug wreck, whose fly-around render accompanies this release, are tuning the training side of the same pipeline.
Below is an edited version of the Charleston Tug, with downscaled photos and with post-processing applied on the point cloud.
Charleston Tug fly-around post-processing render
Charleston Tug - Post Processed"Jake took a pipeline that was leaving nearly a quarter of our photos on the seafloor and turned it into one clean model with only the relevant data included" said Joshua Gross, Director and Manager of Research and Development at Scuba.Tech. "The methods are open and the results are repeatable. What limits us now is compute: with more GPU memory and processing power, we could run full-resolution datasets, all three cameras at once, and get answers in hours instead of weeks."
Computation power is now the program's bottleneck. GPU memory caps force the team to decimate datasets, limited CPU power requires downscaling photos before processing, and each full-scale experiment ties up the lab's donated hardware for a day or more. Scuba.Tech invites collaborators, sponsors, and partners to help expand its compute capacity, deduplicate research, and fund summer and fall fellowships, accelerating native support for triscopic and multiscopic camera inputs across 23Z OSIRIS and 23Z GSHARC. To support this work, visit https://www.scuba.tech/donations or write to hello (at) scuba.tech. Learn more about OSIRIS and GSHARC at https://www.scuba.tech/tools-and-resources/osiris-gsharc.
Scuba.Tech Incorporated is a 501(c)(3) nonprofit research and education lab advancing oceanography, marine biology, and maritime technology through open-source tools, field operations, and student mentorship. We publish methods and media for the public good and collaborate with schools, agencies, and community organizations.
Press & Partnership Contact: hello (at) scuba.tech.
Embedded Media
WORDS FROM THE CREW
Testimonials
Jason C.
Grant writing assistant, Intern"I stay with Scuba.Tech because I believe in the skills I’m learning and the impact they can have on ocean conservation. My goal is to protect and preserve the ocean's splendor while doing great things for our planet.
So far, I’ve contributed to creating animal species guides, supported public ROV dives, and helped with grant writing. With a background in biology and zoology, as well as scuba certification, I’m excited to learn even more—like photogrammetry—and take on bigger roles, such as becoming a science communicator."
Andrew H.
Historian & Composite Review for Stolt 3D, and Grants Writer"Growing up, I was fascinated by shipwrecks, obsessively sketching them and dreaming of visiting famous sites. Thanks to Scuba.Tech, I’ve rekindled that childhood passion and found a unique way to combine it with my skills as a professional historian in the realm of STEM and maritime science.
Scuba.Tech offers a collaborative environment where diverse expertise—from computer engineers to marine biologists—comes together for public education and scientific research. For me, that means crafting the historical narratives of dive sites, shipwrecks, and the impact of human activity on the marine environment. Projects like Stolt 3D allow me to explore the stories beneath the surface, connecting past events to ongoing exploration and conservation efforts.
Maritime exploration isn’t just about visiting underwater destinations—it’s about the memories and connections we bring back to the surface. My advice to aspiring maritime explorers? Tap into your inner adventurer, take chances, and ask around—opportunities are closer than you think!"
Adam S.
R&D Staff, Grant Support"I'm a particle physicist. There are particles in the ocean. That's synergy, Baby! In all seriousness, it is fun to see my friends tackle a world of science, and I am happy to lend my support where I can, whether it be by coming up with crazy project objectives, wordsmithing, or mentoring."
BTS 2025
This presentation was originally given on 03/29/25 at Beneath The Sea 2025; and is now shared here to benefit the scientific & education communities for public good!
Page updated
Google Sites
Report abuse