From kde-devel Thu Mar 08 08:52:44 2012 From: Sune Vuorela Date: Thu, 08 Mar 2012 08:52:44 +0000 To: kde-devel Subject: Re: GSoC idea: improving scanning and OCR in KDE (skanlite/kooka) Message-Id: X-MARC-Message: https://marc.info/?l=kde-devel&m=133119685320896 On 2012-03-07, K=E5re S=E4rs wrote: > How do we go forward from here? = Put it on the wikipage, wait for KDE to be a gsoc approved organization, wait for official project submissions, get the student to submit the project. hope that KDE as a organization selects your project. > Who is willing to mentor for an ORC library + plugins + ...? = My best guess for a great mentor for such a project would be someone with a iki.fi address with a firstname and a lastname of 4 letters, both containing some 'nordic' letters. :) > What does a mentor do and how much time does it take? a mentor answers questions from the student. A mentor ensures the student is on right track. A mentor fails or passes the project. A mentor does a lot of code review A mentor also ensures that the project plan is good and sane. Mentoring, depending on student and the task, is probably 1-2 hours a day or something like that. Fram my experience with santa, he is probably going to be one of the students who is going to be pretty light to mentor. /Sune > > /K=E5re > > On Wednesday 07 March 2012 16:16:59 todd rme wrote: >> 2012/3/7 K=E5re S=E4rs : >> > On Wednesday 07 March 2012 10:59:50 todd rme wrote: >> >> On Wed, Mar 7, 2012 at 10:46 AM, Andreas Pakulat wrote: >> >> > On 07.03.12 10:23:32, todd rme wrote: >> >> >> On Tue, Mar 6, 2012 at 8:03 PM, Klaas Freitag wr= ote: >> >> >> > On 06.03.2012 18:02, todd rme wrote: >> >> > [...] >> >> > = >> >> >> > These kind of things. Not sure if a kio is cool for any of these. >> >> >> = >> >> >> A gui able to do all the things you listed would necessarily be >> >> >> extremely complicated and likely difficult to use, unless most of = the >> >> >> tasks were automated push-button affairs. In the latter case, the= re >> >> >> is little advantage over a kio slave. I would think that a kio sl= ave >> >> >> would be more natural, since users would not need to know terminol= ogy >> >> >> or the menu structure. >> >> > = >> >> > Maybe I didn't use enough of the more fancy kio-slaves, but I have a >> >> > hard time imagining how I'd be able to use this with say konqueror.= I'd >> >> > go to >> >> > = >> >> > kscan:/// >> >> > = >> >> > And then see whats been scanned, but how do I initiate a scan? Do I >> >> > need >> >> > to go to some special url? If so, how do I trigger the OCR creation >> >> > after scanning? >> >> = >> >> To activate a scan of an image, you either drag the image file in the >> >> kio slave to another folder, or you open it in a program (either by >> >> clicking or using the right-click menu). In the case of dragging it >> >> to a folder, it will be automatically scanned and saved in the >> >> destination folder without the user needing to do anything else. In >> >> the case where you open it in a program, it will probably be scanned >> >> to a temporary folder or stored in memory and then opened in the >> >> program, once again without the user doing anything else. >> >> = >> >> In the case of OCR, it would be the same, except a temporary image >> >> file woulds be scanned, OCRed, and deleted (or again stored in >> >> memory). >> >> = >> >> This, at least, is how the CD kio slave does it. >> > = >> > If somebody is interested in making such a kio slave, for simple useca= ses, >> > I would say go ahead and scratch your itch :) I do have a some doubts >> > about the usability tho. >> > = >> > 1) You would have to "refresh" the view to get a new preview of new ph= otos >> > placed on the scanner and the automatic photo finder is bound to fail >> > sometimes and you would be unable to select the correct part of the >> > images. >> = >> Yes, refreshing would be needed, both for this and for a standalone app. >> = >> The issue with incorrectly detected borders would also affect a >> standalone app. Of course this is intended for simple jobs, anything >> complicated would need a more advanced app. But for most cases simple >> is enough. >> = >> > 2) You have options (folders?) >> > - scan mode: grayscale, color >> > - resolution 50 100 150 300 600 1200 2400 4800 ... >> > - source: flatbed, automatic document feeder, transparency unit, ... >> > - how would you adjust gamma if available >> > - contrast/light... >> = >> The only folders would probably be resolution, and one extra folder >> for the ADF if available. The ADF would primarily be useful for PDFs, >> TIFFs, and OCR, so in that folder could be individual files for OCR, >> PDFs at various resolutions, and TIFFs at various resolutions. >> = >> Color vs. grayscale could have two images for the whole scan, so only >> one more file per resolution. OCR would handle that automatically, >> and scanned photos are unlikely to be in grayscale. >> = >> Transparency units usually replace the main scan bed, so the could be >> detected as individual pictures and scanned that way. >> = >> Gamma, contrast, lightness, etc would require a standalone app. >> = >> > 3) Multipage scanning from ADF can not have a preview... >> = >> No, but this is true in a standalone app as well. >> = >> > For simple point and shoot it might work some of the time but I'm not = sure >> > the amount of bug reports for heuristics failures would be fun to go >> > through ;) >> The same bug reports would be needed for a standalone app, since it >> would be using the same defaults. >> = >> -Todd >> = >> >> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubsc= ribe >> >> << > >>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscri= be << > >> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscrib= e <<