bochenys / satellite-pose-estimation

Satellite Pose Estimation with Deep Landmark Regression and Nonlinear Pose Refinement

Satellite Pose Estimation with Deep Landmark Regression and Nonlinear Pose Refinement (ICCVW 2019)

We propose the following pipeline for satellite pose estimation.

Our approach won the 1st place in the Kelvin's Pose Estimation Challenge.

Tha relevant dataset can be downloaded from https://kelvins.esa.int/satellite-pose-estimation-challenge/data/.

Cite this work:

@InProceedings{chen2019satellite,
  title={Satellite Pose Estimation with Deep Landmark Regression and Nonlinear Pose Refinement},
  author={Chen, Bo and Cao, Jiewei and Parra, Alvaro and Chin, Tat-Jun},
  booktitle={ICCVW},
  year={2019}
}

satellite-pose-estimation's Issues

Error Compile openMVG MEX function

Hi, thanks for releasing the code.

When I tried to compile the OpenMVG mex function for Matlab, the errors occured:

/usr/bin/ld: /usr/local/lib/libopenMVG_geometry.a(Similarity3.cpp.o): relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
/usr/local/lib/libopenMVG_geometry.a: error adding symbols: Bad value
collect2: error: ld returned 1 exit status.

Could you please give me some suggestions?

Issure about SIMULATED ANNEALING LEAST SQUARES POSE ESTIMATION

Hi.

In your source code on SA-LMPE experiments, you provide 12000 views predicted 2D keypoints results. I am wondering if you uses these predicted（12000 views）results to estimation single 3D pose for satellite, or these 12000 views are independent without geometry constraint?

Thanks.

Where can I get Error Data

Questions on Setup

Hello,

Thanks for the well written paper / documentation!

I am trying to recreate your results / learn about machine learning packages you used.
For the first step of your paper "Multi-view Triangulation" is the code available on this repository?
The only code on triangulation code I see is for the final step - LMPE.

I browsed through a view .m files but I am not seeing anything that uses "triangulateMultiview" as stated in your paper.

Thank you,
GB