Decimating the Model – the V83

Scapa Processing Continues

What seems like an age ago we were in Scapa Flow diving from the good ship MV Valkyrie. Some of the models were completed on board – like the discovery of a German Pinnace whereas others such as the UB 116 took a little longer.

Near the end of the list of wrecks to process is the V83. Here’s the results:

Some, but not all…

Compared to some of the Scapa wrecks, the V83 is a small site and sits on a slope ranging from 18m at the stern to 4m in the shallows. It is also covered in the photogrammetry nightmare that is brown seaweed. This stuff moves all the time and makes aligning images a challenge.

Changing Profiles

The changing depth of the wreck does not help scanning either. When diving, its a very good idea to reach your maximum depth near the beginning of the dive and then work progressively shallower. Constantly going up and down with depth can help trigger a bend, something we know all about…and are keen to avoid. Careful planning of the order of scanning can help mitigate the risk, but can mean areas are missed, or end up being scanned multiple times.

Missing Bits

All of which goes to explain why we have some missing sections of the wreck. Moving weed and a slightly chaotic site have meant some images just will not align. But even with missing sections we can still appreciate the site layout.

Decimating for Sketchfab

Agisoft Photoscan Pro does an excellent job of trying to fill the gaps by interpolating the mesh. But it can also pick up spurious points when shooting horizontally or slightly vertically in the water column. With the exception of a few places in the world, water has suspended particles and these can be identified as being part of the dense cloud. Here’s an example:

Screen shot of V83 model

Screen shot of the V83 model before the mesh was edited. Note the light sections of the mesh.

So for this published model I have a) decided to trim off the spurious faces and b) reduce the face count down from approximately 2.5 million to something more internet-friendly, hence the two separate chunks in Photoscan:-

Screen shot of the V83 model in Photoscan

Screen shot of the V83 model with two chunks; one is considered master and will have more data added to it. The other is an intermediate publication chunk decimated for Sketchfab


The Sketchfab version can be ditched at some point in the future, but for now we will hang onto it. So just like the F-2 and YC-21 barge it remains work in progress. Unlike the other wrecks the V83 will not be available as a wreck print just yet as the quality is not quite as good as we expect at this stage.