wrong extrinsics about omnimvs_pytorch

[matsuren]

Sorry for late reply. It's been hectic.
Hmm, the last time i checked omnihouse, the locations of the cameras was something like:
cam0:[0.0, 0.0, 0.0],
cam1:[20.0, 0.0, -20.0],
cam2:[0.0, 0.0, -40.0],
cam3:[-20.0, 0.0, -20.0],

And I visualized the camera pose that seemed to be okay.
You might want to see my notebook to visualize the pose here.

Anyway, thank you for your information. The next I use omnihouse, i will be careful!

I'm now working with habitat-sim/habitat-lab, and I'd like to thank you for your implementation of the double sphere model!

Glad that you're using double sphere model 😄

[matsuren]

Thank you for your information. But, I don't know which datasets you're talking about. At least, omnihouse and omnithings datasets seems to be correct for me.

Thanks,

[Siidej]

Hello,
I've tested omihouse, it corresponds to its yaml file. I also tested omnithings, and none of its two files works. I think maybe there are some problems after their updating of datasets.
I'm now working with habitat-sim/habitat-lab, and I'd like to thank you for your implementation of the double sphere model !

[matsuren]

@Siidej
It seems that you're right about the extrinsic parameters. Check the discussion here. The authors of the datasets keep changing so be careful about my source code. The groundtruth depths were provided as png when I downloaded, but it seems that they are provided as tiff.