Rewrite sample corr using torch like RAFT

princeton-vl / DROID-SLAM

BSD 3-Clause "New" or "Revised" License

1.66k stars 273 forks source link

Rewrite sample corr using torch like RAFT #23

Closed HardysJin closed 2 years ago

HardysJin commented 2 years ago

Hi, thanks for the wonderful work.

I got some questions

why use correlation_kernels.cu for corr sampling? Is it faster?
what corr_index_forward_kernel function in correlation_kernels.cu is doing? Sorry I am a little confused.
Is that possible to rewrite the sample corr using torch (or put it in a network format) like what RAFT did?

Thanks

zachteed commented 2 years ago

Hi, yes the corr sampling is exactly the same as RAFT. I implemented it as a cuda kernel to improve speed and decrease memory usage. The grid sampling operation in pytorch was a fairly large bottleneck in raft.

HardysJin commented 2 years ago

Thanks, got it work