Fossil Photogrammetry

 To go with the creation of my museum environment, I thought I should take advantage of person resources available to me. I have a private fossil collection with quite a wide variety and combining that with my own knowledge of palaeontology means I could create a quite professional exhibit. 

Having never used photogrammetry before this was a new experience and I still have some issues to work out but now I have a good grasp on how to do it.

I tried to test various photogrammetry software using a small statuette of Eivor from the Assassin's Creed Valhalla Collector's Edition. After a few hours of testing different software I determined that the object was simply too small for my camera to focus on it properly or the software to detect it due to how close my camera was.

The next day I tried one of my largest fossils to see if size was the issue. Surely enough I was able to create a working cloud generation using both 3DF Zephyr and Meshroom. I prefer using Zephyr due to it's ability to process a video into frames used for mesh generation (50 frames maximum for free).

The dense cloud generation and mesh generation were simple automated processes and for each I chose the close up and deep detail presets to ensure the best result.

Before I got the final mesh below, I had many redos testing the differences between images and videos, playing with the video processing settings, cropping the background and messing with presets.

My inital tests were with the object laying down. When I realised the difficulty this would cause with separating it I chose to place it on the stand. This stand created its own set of problems as removing this would be impossible without leaving a whole in the model.

Regardless, I chose to move into Maya to get a general workflow for photogrammetry.

The original model was almost 1 million tris which I then used the remesh tool on taking it to 3 million tris. After hitting it with the reduce tool, it was brought down to 100K tris which was a more comfortable starting point.

I soon realised that a lot of the details on the model was accounting for the irregularities on the texture of the fossil. So I reduced it by half again knowing that the texture itself would provide a decent representation of the roughness.

Using this new 50k one as a base, reduced it by over half again and then by about 90% to compare the two. I determined that the texture does a fantastic job at covering up the which model had less detail in terms of polys. Because of this I think the 7k model would work great in my collection and is a lot more resource efficient on the engine.

I tried using the retopologize tool to see if could help make the UVs a little easier to understand but because I have little control over the creation of the texture this warped and stretched it as well as making the object look too smooth.

This final experiment involved me manually removing the stand and sewing a duplicated part of the model to it in place of the stand. This I have realised will create too many issues with the texture and was not worth my time continuing.

To fix the issue with the stand and my problem with indirect lighting on the back of the object, I intent to first capture the front and back individually and create two separate meshes. After cutting them both in half, I hope to have two meshes I can merge to create one. The only downside to this is having two textures for the object which I will look into fixing but I think that is a small price to pay for a competed model that can be viewed from all sides.