search

SimonTheVillain / ScalableFusion

Completely mesh based RGB-D reconstruction. Capable of storing large reconstructions without wasting GPU memory.
GNU General Public License v3.0
77 stars 12 forks source link

How to run? #10

Closed zmf2022 closed 1 year ago

zmf2022 commented 1 year ago

I have building success, but do not konw how to run it. it miss much configuration file

SimonTheVillain commented 1 year ago

Honestly, I stopped developing this code long ago but I might be able to help. How are you trying to run the reconstruction? Do you have a TUM style dataset? Did you create the associations.txt?

EDIT: What are you expecting from this reconstruction? Depending on what you want to achieve, I might be able to tell you if and how this is achievable with ScalableFusion. Also sorry for the code-quality in this repo, I was quite inexperienced when starting this project.

zmf2022 commented 1 year ago

Thank you very much, i have downlaod tum datasets  and generating associate.txt, but it seems need something other files,such as rgb_exposure.txt extrinsics.yml?

---Original--- From: @.> Date: Wed, Jan 18, 2023 04:39 AM To: @.>; Cc: @.**@.>; Subject: Re: [SimonTheVillain/ScalableFusion] How to run? (Issue #10)

Honestly, I stopped developing this code long ago but I might be able to help. How are you trying to run the reconstruction? Do you have a TUM style dataset? Did you create the associations.txt?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

SimonTheVillain commented 1 year ago

Thats unfortunate, I hoped that I would have implemented proper fallbacks in case no configs are provided. Try this file as intrinsics file and put it in the folder above your your sequence extrinsics.zip The rgb_exposure.txt looks something like this: 30 rgb/000000.004281.png 30 rgb/000000.028649.png ... Its probably best to write a small python script to generate this sequence for you. The first number is the exposure time used during capture. It probably was not constant during TUMs capture process, but it was not logged either. It thus is best to just set it constant. In case you just want to check out any sequence, download this. If even that sequence doesn't work, I have made a few changes locally that I might need to push to master for you to run this.

It would still be nice if you could tell me what you are expecting from this reconstruction. I have a good grasp on what it does and what it doesn't do. I hope it is not your professor, boss or somebody else forcing you to get this code to work.