DLL is lost when running the binary

[xianggao1989]

Hi,
I want to use your released binary CODE.exe. However, I was warned that the nppc64_70.dll is lost in the computer which I can not find in your binary folder WideBaselineFeatureMatcher_PAMI-master/CODE/. How can I deal with this?
Thanks a lot!
Xiang Gao from NLPR, CASIA

[xianggao1989]

Hi, I have known where is wrong (my CUDA version is not 7.0) and now the .exe can run. I have another question, in the README.md it is said the commands "-pwis" will take a .txt file as input and perform full matching between all pairs of images in the input list and "-full" will take two .txt files as inputs and perform full matching between every image in the first list and every image in the second list. I wander if I put one of them in my command line, the resulted point matches have been filtered or not? I am not sure about that because the command '-blat' can not be put explicitly in the command line.

[seravee08]

Hi, Yes, all these options will lead to filtered matches regardless of '-blat' present or not.

…

On Wed, Dec 6, 2017 at 7:50 AM, xgao_nlpr ***@***.***> wrote: Hi, I have known where is wrong (my CUDA version is not 7.0) and now the .exe can run. I have another question, in the README.md it is said the commands "-pwis" will take a .txt file as input and perform full matching between all pairs of images in the input list and "-full" will take two .txt files as inputs and perform full matching between every image in the first list and every image in the second list. I wander if I put one of them in my command line, the resulted point matches have been filtered or not? I am not sure about that because the command '-blat' can not be put explicitly in the command line. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <https://mailtrack.io/trace/link/f682e349abcf20be63767bdfaa3d33736032b5bf?url=https%3A%2F%2Fgithub.com%2Fseravee08%2FWideBaselineFeatureMatcher_PAMI%2Fissues%2F2%23issuecomment-349630369&userId=622087&signature=4da3490eba9985f6>, or mute the thread <https://mailtrack.io/trace/link/7f224e7cd59198fd97006ee8575bff451703dd23?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAFR55ALnuZPKgNTdFkv553iiE6E-Ollyks5s9o2kgaJpZM4Q3veo&userId=622087&signature=9358756bf73432d0> .

[xianggao1989]

Thanks for your reply!
I meet another problem now. In the README.md file, -noaf means no affine. I put it in my command line but the outputs of the feature detection still contain two kinds of files, .sift and .aff, and the sizes of the files are the same whether the -noaf is in the command line or not. I am not sure whether the command -noaf works well. Can you help me with that?
Many thanks!

[xianggao1989]

Still another question. Can I run the binary with my own features and descriptors? The output of the feature detection are .sift and .aff files, both in binary. I guess the .sift file has the same format with those of VSfM, though I am not quite sure about that. Can you affirm my guess and tell me the format of the .aff files?
Many many thanks again!!