On May 22, 2011 01:43:01 PM kfj wrote: > On 22 Mai, 16:10, Yuval Levy <goo...@levy.ch> wrote: > > that's theory. What I observe in practice are curled rows that taint the > > rest of the strategy after the rows have been identified and optimized > > first time. > > maybe the curled stripes react well to the horizon straightening > routine. Just an idea.
the problem is that at this point it's already too late and I am in manual mode. So I might as well select manually the image pairs and run cpfind each time on them. If there was a button to trigger cpfind on the cp tab, this would be an easy way to waste some time clicking away... ... but I can also try to script this in python, can I? > I do a 360x180 with a fisheye as a backdrop and then 'pin' the longer- > lens shots to it. I did not have the luxury. There was no fisheye anywhere near me. But yes, in ideal circumstances, that's the smart thing to do. > I must admit that I hadn't looked into the cpfind documentation fore a > while, and that the text that is now in the wiki is actually good and > helpful. I often simply call the program in question, or call man with > it, and what you get this way is pretty thin. Sorry for being sloppy. I skimmed over the wiki page and it looks suspiciously similar to the man page. > I would dearly like to understand more of how cpfind is doing what it > does. I made several attempts to look into the code but I find it hard > to penetrate. Yes, I found the code hard to navigate too. When I submitted my patch this morning, I just peeled away at it onion layer after onion layer starting from the CLI switches and did not get much far - just enough for a small hack. The frustrating thing with this is that if I focus hard and apply myself, I might get to understand it and do some things with it in a few hours. But two weeks down the road this knowledge is already forgotten and I am back to square zero trying to understand things that I understood weeks earlier but that have gone out of my mind in the meantime. > hsi/hpi is just a first step in the right direction. The helper > programs (CPGs, warping, blending) can all be made into modules with a > bit of SWIG magic we need a list of functionalities already available; a todo-list of functionality to expose next; a how-to for doing it. > > Good news for Ubuntu users: I updated the wiki instructions and now > > everybody with Lucid (10.04) and later can access python scripting; and > > Philipp updated the nightlies build process so that now the python > > scripting interface is available to those brave enough to type the > > following commands in the CLI: > > > > sudo add-apt-repository ppa:hugin/hugin-builds > > sudo add-apt-repository ppa:hugin/nightly > > sudo apt-get update > > sudo apt-get install hugin > > this sounds exciting! Does it mean that the hugin you get from the > PPAs is now per default an hsi/hpi-enabled version? Short answer: YES. Long answer: there are two sets of binaries in the PPA: "nightlies" and "builds". At this very moment the unconditional YES applies to the nightlies only. Hugin-builds is for builds from tarball and the YES will apply to hugin-builds shortly after the release of 2011.2.0_beta1. Users not fearing the bleeding edge use the nightlies and have full access now. Conclusion: we want to move fast on releasing 2011.2.0. Yuv
signature.asc
Description: This is a digitally signed message part.