General Discussion: Mapping a Video Game City and Memory

Howdy OpenDroneMap gurus!

I joined a cool game enthusiast group and we’re mapping different parts of an online game called Star Citizen. It includes full planets to explore (albeit at a fraction of the scale of Earth) and I’m finding it quite fun to try to create maps via OpenDroneMap. So far it has been great learning more about good drone mapping techniques whilst pretending my space ship is a “drone”. I’ll have to invest in a real drone one day. :smiley:

Fun challenges when mapping in a game:

  • Everything is manual, no path programming
  • GPS data is lacking (although I found a way to inject some Exif info like altitude and focal length)
  • 3D draw distance and level of detail at different distances creates an interesting challenge.
  • Days are MUCH shorter, making shadows an issue possibly.
  • Landmarks tend to be rather epic.

Here’s a quick small example:

(soon, I can only add one image at current privilege, will fill in with link shortly)

and a larger example I started before realizing I needed to start smaller…

(soon, I can only add one image at current privilege, will fill in with link shortly)

Here’s an example for today’s discussion (total RAM hog I LOVE it!):

hurston-ortho

and it’s camera plots (maybe good for discussion as well with cams way off of orthophoto?)

(soon, I can only add one image at current privilege, will fill in with link shortly)

In that last example, it works pretty good as a “Fast Orthophoto” preset on Windows 10 Pro Docker Install with 16GB RAM used, 221 images. However, I could use a bit of guidance of what to look for to determine RAM requirements when scaling up detail. A very detailed orthophoto map would be cool!

I can follow up with more detail if needed, yet it seemed like I’m running into RAM issues based previous searches when adding DSM or DMT maps, or scaling things up in general (mesh size, ortho resolution, etc.). I either get an empty point cloud error or pdal tasks running forever at very low/no CPU (both point to memory, sorry for lack of logs at the moment).

AWS has great options with up to 996GB RAM at a glance. Wow! Expensive at that size too… Finding the happy ground based on any errors could help. Am I able to see something that would point to needs if I enable debug mode? Any tips on how to find that info while trying out different settings? Thanks for the feedback while not providing exact logs at this time. It’s been more of a learning/art-like experience thus far with no huge expectations at the start. My personal computer would disagree at this point though lol!

Thank you for what you do. OpenDroneMap rocks!!

4 Likes

Welcome!

This is absolutely astounding stuff!

I love the intersection of GIS/Remote Sensing and gaming, and you’re doing some really neat stuff in this space!

Pretty awesome that you were able to extract camera position/world position from the game engine to inject into the photos!

RAM limitations are pretty hard to suss out for me at least. I do know that feature quality and min-num-features have a HUGE impact, especially early on in the processing pipeline. I’m not sure how far you can push that dataset with 16GB, likely with 12GB or less actually free at a given time.

1 Like

Thanks! It sure is fun.

Good to know on the RAM. Thanks for helping to set expectations there. I’ll have to play with those values you mentioned.

I was able to launch an AWS ubuntu instance with 64GB RAM having same results. It’s possible that I had those values set high though and need to scale up even more, itching to try an instance with more CPUs and perhaps 256GB of RAM next. I’ll do this after another couple of photo runs with updated techniques. :slight_smile:

1 Like

Moar RAM only ever hurts our pockets, never the reconstruction :wink:

Fun is good. Play is important as it allows us to interact with things differently and push things beyond where they are.

Please, do keep experimenting and letting us know how it goes!

Can you link into other programs/engines?