search

zjuluolun / BEVPlace

A rotation-invariant, generalizable, and SOTA LiDAR place recognition method.
132 stars 8 forks source link

Registration on BEV #10

Open Snow-know-nothing opened 2 weeks ago

Snow-know-nothing commented 2 weeks ago

Thank you for your work! I am wondering if I can use the local features output by the group convolutional network for registration between BEV images to get 2D pose?

zjuluolun commented 2 weeks ago

Thanks for your interest. Yes, registration on BEV can produce accurate 2D poses. I'm preparing a journal extension of BEVPlace, which has included this idea. It seems that we are on the same page hh!

Snow-know-nothing commented 2 weeks ago

Wow，hope to see it soon！