search

ubc-vision / COTR

Code release for "COTR: Correspondence Transformer for Matching Across Images"(ICCV 2021)
Apache License 2.0
460 stars 58 forks source link

About HPatches datasets #34

Open ifuramango opened 2 years ago

ifuramango commented 2 years ago

Thanks very much for your great work! AND i want to know that how do you test and evaluate the HPatches dataset(in the code)? Can you tell me how to get the relevant code?

ifuramango commented 2 years ago

Since I'm new to feature matching, I'd be very grateful if you could help!!!

jiangwei221 commented 2 years ago

Hey, we evaluate HPatches with the COTR sparse engine. It will provide sparse correspondences, and we will densify/interpolate the correspondence values. The final resulting dense correspondences area will be slightly smaller than the GT mask, because some corner/edge correspondences will be filtered out. After the data is dumped, we use a HPatch benchmark to evaluate the AEPE. We have the untrimmed code to run COTR on HPatches and the benchmark code, if you want the code, please send me an email at: jiangwei1993@gmail.com

shy922 commented 1 year ago

Since I'm new to feature matching, I'd be very grateful if you could help!!!

I'm a new to feature matching, too. Can we talk to each other?