This is a very short post to announce a SFMToolkit bug fix… If the toolkit wasn’t working at all on your machine (File not found error in “1 – Bundle.wsf”) this may fix the issue. The bug was linked to the Windows separator setting: ‘.‘ or ‘,‘. Thus the default 0.8 matching threshold was on some system applied as 0: no matching -> no bundler output 
I’ve also fixed other errors in BundlerMatcher too (small memory leak + no more 4096 match limit), so you should download this version too even if the previous version was working on your system. The new version is available on his dedicated page: please do not make direct link to the zip file but to this page so people downloading will always get the latest version.
Thanks Henri! Just had it work on a bundle that I have been trying to run for a few weeks!
@Greg: sorry for the long bug fixing delay! I just have re-installed my computer and the toolkit wasn’t working because of this Windows setting… Fixing bug is always faster when it happens on your machine
Hi there. So Ati users should have problems using this kit? because I’m having them and dont know if they’re related to this. I unzip the kit in C. Copy images in the samples directory generate list…. and while runing sparse recon the process starts, each image is “compared” with the others and at the end it says that 0 good images had been found. But with these images I’ve been able build using photosynth + PMVS quite nice point clouds…. Also if I set the “resolution” as 0 I’ve the feeling that the app never starts to compute the stuff.
I get the following error
error message windows title SIFT_GPU: BundlerMatcher.exe – Applicaion error.
The instruction in “0x690ca897″ makes reference to the memory in “0×00000010″. The memory cannot be “read”.
I’m still getting the same “File not found error in “1 – Bundle.wsf”” error with this update. Any clues to why?