search

patr-schm / surface-maps-via-adaptive-triangulations

Implementation of the paper "Surface Maps via Adaptive Triangulations" (Eurographics 2023)
https://www.graphics.rwth-aachen.de/publication/03345/
MIT License
44 stars 3 forks source link

n_meshes demo slow after set distortion pairs #2

Open wchengad opened 9 months ago

wchengad commented 9 months ago

Hi Patrick,

Thanks a lot for the tremendous serious of work, I was running your code and trying to get the 'spot and friends' result, and I thought it might be very similar to the process like n_meshes, just inserting 'spot' as the first reference mesh and replace the texture map, right?

But the thing is, it runs smoothly before set distortion pairs, then it is stuck there, with no log output in 30 minutes, and the memory usage is still there, I don't know what might be the bug. My coarse_to_fine demo runs perfectly smooth, btw.

I also tried to reduce the pair number, saying '{381, 382, 383, 384, 385}', but it still paused there.

patr-schm commented 8 months ago

Hi, thanks for the nice words! I don't remember this issue. Can you try computing a pairwise maps from spot to the others first? Also, you could double-check if the landmarks you're using are actually compatible. And maybe check that there are no wrong sphere embeddings in the output folder that were cached in previous attempts.