When I first started diving the UB88, I was also starting to investigate and learn photogrammetry. At the time (about a year ago), I thought “wow, building a model of the UB88 would be a cool project.”
I’ve been posting updates as I’ve been working on the project over the past year.
Doing a large scale model on a deep wreck can present challenges, especially when the water visibility is often limited. There is generally a tradeoff between time and detail. With a scooter and a GoPro, an experienced photogrammetry person could probably build a model in a single dive. The model would give a good idea of the wreck but wouldn’t be very detailed. I took the approach of taking more detailed pictures of a specific section of the wreck on each dive and then piecing it all together.
The result is a more detailed model but it took me quite a few dives (I provided a summary of each dive near the end).
The model was built on a 2017 iMac Pro computer with a 3.2 GHz 8-Core Intel Xeon W CPU, a Radeon Pro Vega 56 8 GB graphics card, and 32 GB of DDR4 memory. It consists of 2472 photos. Here are some key statistics for the “medium” quality model:
- Image Alignment : 12-15 hours (estimated – completed in stages)
- Depth Maps : 1 hour
- Dense Cloud : 3 hours
- 3D Model : 5 hours
- 3D Model Faces : 11M
- 3D Model Vertices : 5.5M
Complete UB88 Model and Screen Captures
Below is a link to the low-res version of the model that is available for free on Sketchfab (low res for this size model was required to fit within the upload size constraints).
There are two issues with this model:
The first is the lack of a consistent coloring. This is due to primarily two different factors. The first factor is that the color of the water changes in California quite noticeably depending upon conditions and season. I’ve seen it downright murky green and I’ve seen it blue. For the most part, I didn’t use any tools such as Lightroom to try to create a consistent color tint. The second factor is the type of lightning used. I switched from using a Big Blue VTL light which has a very “hot” or “white” temperature to a Big Blue CB light which has a “warmer” or less white tint. You can see the “white” light on the front port side torpedo tubes.
The second is the lack of the ghost net and the very end of the stern torpedo tube. I have mentioned this in a few other posts and I do have hopes of adding that at some point in the future but I didn’t want to wait for that before publishing the model.
Here are the top, starboard, and port side views of the model:
Detailed Screen Captures
I’ve also included multiple screen captures of some of the more interesting areas of the wreck.
Dive / Project Summary and History
For anybody interested, below is a table of the dives, dates, and results of each one. I’ve learned a lot of lessons on this project and could probably do it in half the number of dives and get better results.
|Dive #||Date||Objective||Notes / Results / Screenshots / Blog Post Links|
|1||29 Oct 2020||Forward Torpedo Tubes||Blog Post Link|
Port and Starboard tube models but not connected together. Port side had 162 photos and starboard side had 56 photos
|2||29 Oct 2020||Conning Tower||Blog Post Link|
144 photos. Model incomplete. Learned that getting the top or end of something is hard. In this case, getting the top of the periscope tubes was challenging.
|3||25 Nov 2020||Finish Forward Torpedo Tubes / Finish Conning Tower||Blog Post Link– Torpedo Tubes|
Blog Post Link – Conning Tower
I was able to connect the port and starboard side torpedo tubes to build a complete model of the front section of the wreck. I also was able to capture/connect the small debris field out in front of the torpedo tubes.
The conning tower also came out well but I was not able to connect the conning tower and top deck to the torpedo tubes so I had two big sections done but not the hull on either side.
|4||25 Nov 2020||Stern Torpedo Tube & Ghost Net||Blog Post Link|
This one was a hard lesson. It was my first attempt at getting a “good” model of the torpedo tube at the stern and the ghost net (I recently had a second attempt that also didn’t work the way I had planned.
|5||9 Sep 2021||Finish the entire front section of the model||Blog Post Link|
Justin and I tried a UB88 dive on 10 Dec and the visibility was so bad that we couldn’t even find the wreck even though it is a 190′ long tube. It was a loooong time before I got back to the UB88 after that.
The dive went really well. I was able to get enough pictures of both sides of the hull in a single dive to stitch together a model of the entire front half of the sub.
|6 & 7||18 Sep 2021||Build both sides of the rear section behind the conning tower.||Blog Post Link|
I spent the first dive on the port side and the second dive on the starboard side. My plan was to get enough of the conning tower to connect it to the front half of the wreck.
There results were good and I was able to build a model of the entire back half of the sub with just two dives.
|8||14 Oct 2021||Stern torpedo tube & ghost net (take 2)||Blog Post Link|
Another dive on the stern to try to finish the model, including the very back end of the stern torpedo tube. I had tried this once before and planned a second attempt. It ended similar to the first attempt.
First, a big thanks to Ray & Kyaa at Sundiver for putting me on the UB88 the first time and each additional dive. They never miss the drop and are always helpful.
Second, thanks to all of my dive partners. Ben Lair helped organize the trip with Sundiver to get on the UB88 the first time back in July 2020 and helped me extract myself when I did my interior penetration dive a month later. Drew Wilson and I have done many, many dives together on the sub working on this project and you should really check out his UB88 ambient light photos. I’ve also spent a bunch of time with Justin Judd and DJ Mansfield on UB88 dives. I hope I didn’t forget anybody.
I’m considering the project primarily complete at this point. I do have a few things I’d like to work on as time permits, but they aren’t on my “official” project list.
- Human Model – To give a sense of scale, I’d like to include a diver in the model
- Ghost Net – I will, at some point, figure out how to do this. I’m very close
- Mosaic – I’d like to create a new version of the “top-down” mosaic that the UB88 team did. I have tried this before and I know how to do it very well using photogrammetry to increase the accuracy. I just need to do another dive with a partner to get the correct illumination.