search

NVlabs / BundleSDF

[CVPR 2023] BundleSDF: Neural 6-DoF Tracking and 3D Reconstruction of Unknown Objects
https://bundlesdf.github.io/
Other
1.07k stars 118 forks source link

CANT COMPREHEND README! HOW TO RUN THE CONTAINER ? #131

Closed 52THANOS closed 10 months ago

52THANOS commented 10 months ago

after i built a container using the image , the container shows it is running,but i cant bash build.sh. it seems the environment of this container is wrong. image image it shows that im in docker-desktop ,and i cant find my bundlesdf file. im confused. do i have to copy bundlesdf into this container?

52THANOS commented 10 months ago

what directory structure it is

fedona commented 10 months ago

Once you activated the docker image through the bash script that is provided, you then have to navigate to the location of the BundleSDF folder yourself, I do not know where you have it, but start going to /local/home/ might be a good idea.

Once you are in the correct folder you should be able to see the same files that are in this repository and to run the milk example.

52THANOS commented 10 months ago

Once you activated the docker image through the bash script that is provided, you then have to navigate to the location of the BundleSDF folder yourself, I do not know where you have it, but start going to /local/home/ might be a good idea.

Once you are in the correct folder you should be able to see the same files that are in this repository and to run the milk example.

what if i want to run the RGBD video. you guys did not supply this script

fedona commented 10 months ago

I recommend you to first run the milk example and then eventually try to run your own RGBD datasets.

For you own dataset you must be sure the RGB and D images have same camera calibration and resolution. Do not forget to include camera calibration file and to organize datas as in the milk example folder.

All necessary commands are given in the readme file.

p.s. I am not author in this project but I am sure they will be glad to help you, just note that caps lock is somehow a bit intimidating to use.

kevinDrawn commented 10 months ago

Is it possible to use algorithms with real-time camera images?

wenbowen123 commented 10 months ago

@fedona is correct, once you enter docker container, you need to go to the dir of the codebase, which is the same as on your local.

wenbowen123 commented 10 months ago

@kevinDrawn the repo does not support that. It's recommended to record the RGBD video and let it run just like the demo examples. The main goal here is to reconstruct a textured model. Then for pose estimation, we have a recent project, FoundationPose, that can run in real time.

kevinDrawn commented 10 months ago

Wow! Great!! Is this a complete project? If not, when can I utilize it?

wenbowen123 commented 10 months ago

@kevinDrawn It's completed but under review. We will release the code soon after it's accepted.

52THANOS commented 10 months ago

solved

kevinDrawn commented 10 months ago

Wow Thank you so much! I will utilize your new algorithm to make my research!