Re: [Gimp-developer] intelligent scissors
On , 15 Jan 2002, Michael Natterer wrote: > I stripped out the USE_LAPLACIAN part not because it was unused > but because my impression was that it could not work: > > gimp-1-2/app/iscissors.c has the following piece of code: > > #ifdef USE_LAPLACIAN > static gint laplacian [9] = > { > -1, -1, -1, > -1, 8, -1, > -1, -1, -1, > }; > #endif > > but then "USE_LAPLACIAN" never appears below and the "laplacian" > matrix is never used, which made me think it's b0rken, so I removed > it when porting it to GimpTool... Ah ok - it was a long time since I last looked at the code. You're right, it wouldn't work. I mis-remembered there being a bit more to it than that. > I can put the stuff back so people don't need to compare code > they have not restructured themselves, please let me know. Nah: It's not worth the hassle, but thanks for offering. Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
Austin Donnelly <[EMAIL PROTECTED]> writes: > 1.2.2 definitely has the USE_LAPLACIAN code in, but it is not enabled. > > The current developer head in CVS doesn't. I'm assuming that as part > of porting iscissors.c to the new tool model someone (Mitch/Sven?) > took the opportunity to strip out unused code. This is a perfectly > reasonable thing to do. > > However, if Martin wants to compare the performance of the full > Laplacian versus the basic gradient convolutions we do at present, > he'll need to use the 1.2.2 code. Or port that code to the current > developer version. > > In any case, anyone willing to work on iscissors is more than welcome > to it! I agree with Sven that work should use the latest developer > CVS though. Hi Austin, I stripped out the USE_LAPLACIAN part not because it was unused but because my impression was that it could not work: gimp-1-2/app/iscissors.c has the following piece of code: #ifdef USE_LAPLACIAN static gint laplacian [9] = { -1, -1, -1, -1, 8, -1, -1, -1, -1, }; #endif but then "USE_LAPLACIAN" never appears below and the "laplacian" matrix is never used, which made me think it's b0rken, so I removed it when porting it to GimpTool... I can put the stuff back so people don't need to compare code they have not restructured themselves, please let me know. ciao, --Mitch ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
1.2.2 definitely has the USE_LAPLACIAN code in, but it is not enabled. The current developer head in CVS doesn't. I'm assuming that as part of porting iscissors.c to the new tool model someone (Mitch/Sven?) took the opportunity to strip out unused code. This is a perfectly reasonable thing to do. However, if Martin wants to compare the performance of the full Laplacian versus the basic gradient convolutions we do at present, he'll need to use the 1.2.2 code. Or port that code to the current developer version. In any case, anyone willing to work on iscissors is more than welcome to it! I agree with Sven that work should use the latest developer CVS though. Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
Hi, Austin Donnelly <[EMAIL PROTECTED]> writes: > On Tuesday, 15 Jan 2002, Martin Waitz wrote: > > > On Tue, Jan 15, 2002 at 11:08:28AM +, Austin Donnelly wrote: > > > I did test-implement it. #define USE_LAPLACIAN and recompile to see > > > if it makes much difference. I couldn't see a visual different, only > > > it was a bit slower. > > well, then your implementation was not committed to cvs > > It most certainly was. I think someone stripped out that code when > 1.2.2 was re-arranged into the current developer version in CVS. > > Try looking at 1.2.2 app/iscissors.c instead. The 1.2 ChangeLog doesn't show any changes to iscissors.c at all. Is is unlikely that someone changed stuff in 1.2 w/o providing a ChangeLog entry and w/o having me notice it. If it happened nevertheless I'm upset now ;-) The ChangeLog in the HEAD branch shows a couple of changes and indeed all development should happen here. So please, keep looking at gimpiscissorstool.c. If there are patches that we didn't apply, please resubmit them. I will however not apply any undiscussed changes to IScissors since I don't understand the code. I'll do my best however to help anyone with GIMP CVS and if someone wants to take over maintainance of the IScissors tool that would be very much appreciated. We can even arrange to give CVS write access if it helps development. Salut, Sven ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
On Tuesday, 15 Jan 2002, Martin Waitz wrote: > On Tue, Jan 15, 2002 at 11:08:28AM +, Austin Donnelly wrote: > > I did test-implement it. #define USE_LAPLACIAN and recompile to see > > if it makes much difference. I couldn't see a visual different, only > > it was a bit slower. > well, then your implementation was not committed to cvs It most certainly was. I think someone stripped out that code when 1.2.2 was re-arranged into the current developer version in CVS. Try looking at 1.2.2 app/iscissors.c instead. Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
On Tue, Jan 15, 2002 at 11:08:28AM +, Austin Donnelly wrote: > I did test-implement it. #define USE_LAPLACIAN and recompile to see > if it makes much difference. I couldn't see a visual different, only > it was a bit slower. well, then your implementation was not committed to cvs > I haven't looked at the current developer version, so I've no idea > about the path cooling stuff someone else has implemented in the > meantime. i haven't seen anything like this in gimp, too. -- CU, / Friedrich-Alexander University Erlangen, Germany Martin Waitz// [Tali on IRCnet] [tali.home.pages.de] _ __/// - - - - - - - - - - - - - - - - - - - - /// dies ist eine manuell generierte mail, sie beinhaltet// tippfehler und ist auch ohne grossbuchstaben gueltig. / - Wer bereit ist, grundlegende Freiheiten aufzugeben, um sich kurzfristige Sicherheit zu verschaffen, der hat weder Freiheit noch Sicherheit verdient. Benjamin Franklin (1706 - 1790) msg01720/pgp0.pgp Description: PGP signature
Re: [Gimp-developer] intelligent scissors
On Tuesday, 15 Jan 2002, Martin Waitz wrote: > On Mon, Jan 14, 2002 at 10:37:13PM +, Austin Donnelly wrote: > > I think I was the last person to substantially modify the iscissors > > code. I read the paper you reference above, and indeed it was quite > > helpful. However, note that their algorithm is substantially more > > computationally expensive than what iscissors.c currently does. This > > is deliberate. > yes, but i guess one has to test-implement it to be able to > judge if it's worth it... I did test-implement it. #define USE_LAPLACIAN and recompile to see if it makes much difference. I couldn't see a visual different, only it was a bit slower. I haven't looked at the current developer version, so I've no idea about the path cooling stuff someone else has implemented in the meantime. > although i must admit i don't yet understand everything of > gimpiscissortool.c) I don't think anyone does! Good luck! Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
On Mon, Jan 14, 2002 at 03:04:55PM -0800, Laramie Leavitt wrote: > Portions of the 'live-wire' feature would be expensive, namely > the freezing, edge learning, and automatic node generation. what is the difference between 'freezing' and 'automatic node generation' here? both belog to what the paper calls 'path cooling', right? > I submitted a patch not to long ago that implemented the most visible > part of the live wire feature--the interactive edge--and I was told > that it was accepted into the developers version. I have not checked > to see if it really is there, but I could send the file again if I can > find it. yes, it's included in cvs gimp > As far as I can tell, the node generation is not that substantial--it is > easily done by hand. The learning portion would be tough, and with the > current interactive algorithm the freezing would also be difficult. i'll have a deeper look into gimp to try to figure out the problems... -- CU, / Friedrich-Alexander University Erlangen, Germany Martin Waitz// [Tali on IRCnet] [tali.home.pages.de] _ __/// - - - - - - - - - - - - - - - - - - - - /// dies ist eine manuell generierte mail, sie beinhaltet// tippfehler und ist auch ohne grossbuchstaben gueltig. / - Wer bereit ist, grundlegende Freiheiten aufzugeben, um sich kurzfristige Sicherheit zu verschaffen, der hat weder Freiheit noch Sicherheit verdient. Benjamin Franklin (1706 - 1790) msg01716/pgp0.pgp Description: PGP signature
Re: [Gimp-developer] intelligent scissors
On Mon, Jan 14, 2002 at 10:37:13PM +, Austin Donnelly wrote: > I think I was the last person to substantially modify the iscissors > code. I read the paper you reference above, and indeed it was quite > helpful. However, note that their algorithm is substantially more > computationally expensive than what iscissors.c currently does. This > is deliberate. yes, but i guess one has to test-implement it to be able to judge if it's worth it... > Also, the "live wire" feature mentionned in the paper > would be fairly tough to do in gimp (from a UI point of view). this is my problem: i don't know very much about gimp's UI-programming as of now. > Can you be more specific about what features described in the paper > you wanted to add to iscissors? i'd really like to add path-cooling first, as i think it's fairly independent to the underlaying path calculation (as this is done different in gimp from what i've seen, although i must admit i don't yet understand everything of gimpiscissortool.c) i'll read and think on... ;) -- CU, / Friedrich-Alexander University Erlangen, Germany Martin Waitz// [Tali on IRCnet] [tali.home.pages.de] _ __/// - - - - - - - - - - - - - - - - - - - - /// dies ist eine manuell generierte mail, sie beinhaltet// tippfehler und ist auch ohne grossbuchstaben gueltig. / - Wer bereit ist, grundlegende Freiheiten aufzugeben, um sich kurzfristige Sicherheit zu verschaffen, der hat weder Freiheit noch Sicherheit verdient. Benjamin Franklin (1706 - 1790) msg01715/pgp0.pgp Description: PGP signature
RE: [Gimp-developer] intelligent scissors
Portions of the 'live-wire' feature would be expensive, namely the freezing, edge learning, and automatic node generation. I submitted a patch not to long ago that implemented the most visible part of the live wire feature--the interactive edge--and I was told that it was accepted into the developers version. I have not checked to see if it really is there, but I could send the file again if I can find it. As far as I can tell, the node generation is not that substantial--it is easily done by hand. The learning portion would be tough, and with the current interactive algorithm the freezing would also be difficult. Laramie Leavitt -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Austin Donnelly Sent: Monday, January 14, 2002 2:37 PM To: Martin Waitz Cc: [EMAIL PROTECTED] Subject: Re: [Gimp-developer] intelligent scissors On Monday, 14 Jan 2002, Martin Waitz wrote: > anyone working at the iscissor tool at the moment? > > i would like to implement some of the features described in > http://www.cs.orst.edu/~enm/publications/GMIP_98/seg_scissors.html > as a work for an university course. I think I was the last person to substantially modify the iscissors code. I read the paper you reference above, and indeed it was quite helpful. However, note that their algorithm is substantially more computationally expensive than what iscissors.c currently does. This is deliberate. Also, the "live wire" feature mentionned in the paper would be fairly tough to do in gimp (from a UI point of view). Can you be more specific about what features described in the paper you wanted to add to iscissors? Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] intelligent scissors
On Monday, 14 Jan 2002, Martin Waitz wrote: > anyone working at the iscissor tool at the moment? > > i would like to implement some of the features described in > http://www.cs.orst.edu/~enm/publications/GMIP_98/seg_scissors.html > as a work for an university course. I think I was the last person to substantially modify the iscissors code. I read the paper you reference above, and indeed it was quite helpful. However, note that their algorithm is substantially more computationally expensive than what iscissors.c currently does. This is deliberate. Also, the "live wire" feature mentionned in the paper would be fairly tough to do in gimp (from a UI point of view). Can you be more specific about what features described in the paper you wanted to add to iscissors? Austin ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
[Gimp-developer] intelligent scissors
hi, anyone working at the iscissor tool at the moment? i would like to implement some of the features described in http://www.cs.orst.edu/~enm/publications/GMIP_98/seg_scissors.html as a work for an university course. i don't want to duplicate work, so please tell me if there are patches pending for gimpiscissortool.c. i hope i will be able to provide some patches in a few weeks, and will probably have some questions in that time ;) -- CU, / Friedrich-Alexander University Erlangen, Germany Martin Waitz// [Tali on IRCnet] [tali.home.pages.de] _ __/// - - - - - - - - - - - - - - - - - - - - /// dies ist eine manuell generierte mail, sie beinhaltet// tippfehler und ist auch ohne grossbuchstaben gueltig. / - Wer bereit ist, grundlegende Freiheiten aufzugeben, um sich kurzfristige Sicherheit zu verschaffen, der hat weder Freiheit noch Sicherheit verdient. Benjamin Franklin (1706 - 1790) msg01711/pgp0.pgp Description: PGP signature