What I'm doing wrong with face detection?
Locked
What I'm doing wrong with face detection?
 
|
Hi,
I think that I'm doing something wrong with face detection using Digikam 2.0.0-beta5. Just after update to the new beta version, I tried to recognize some faces in my photos. I made a first scan to "Detect and recognize faces". This scan takes about 12hs. After that, I marked some people (about 10), including me, and tried to recognize faces. The result was quite disappointing, digikam suggested to mark with my tag, a lot of faces from people that really doesn't look like me. I'm a 28years old guy from Brazil and it suggested 80y old ladies, dogs, some japanese friends, children, and go on. At first I thought that the algorithm needed more information to be accurate and I confirmed a lot of pictures of me (about 200) and created others tags to mark people that it confuses with me with no success. I made a lot of scans, merged results, tried to recognize faces again, set the accuracy to 100, cleared unconfirmed results and rescan, confirmed about 500 photos manually and I still can't see the face recognition match faces correctly. What I'm doing wrong? Cheers -- Erick Moreno https://profiles.google.com/erickmoreno _______________________________________________ Digikam-users mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-users |
|
|
Hi Erick, Not very helpfull but it may be not
you who is to blame. Rinus Op 20-05-11 03:07, Erick Moreno schreef: Hi, _______________________________________________ Digikam-users mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-users |
Re: What I'm doing wrong with face detection?
|
|
Hi Rinus,
I noticed that too, this happens when you are tagging faces. But, when you try to match new faces with existent tags, the face detection suggest new faces for that tag and put the suggestions together with the faces already confirmed. I see some problems in the way that Digikam is doing this. 1- Of course the first problem is that Digikam can't make good suggestions 2- There is not a visual differentiation between faces already confirmed and faces suggested by digikam and not confirmed yet. Because of it you need to move the cursor over the picture to figure out if it is already confirmed or not. Nobody else think that this is a problem? Cheers Erick Moreno On Fri, May 20, 2011 at 6:23 AM, sleepless <[hidden email]> wrote:
-- Erick Moreno https://profiles.google.com/erickmoreno _______________________________________________ Digikam-users mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-users |
Re: What I'm doing wrong with face detection?
|
|
Am Montag, 23. Mai 2011, 12:00:32 schrieb Erick Moreno:
> Hi Rinus, > > I noticed that too, this happens when you are tagging faces. But, when you > try to match new faces with existent tags, the face detection suggest new > faces for that tag and put the suggestions together with the faces already > confirmed. I see some problems in the way that Digikam is doing this. > > 1- Of course the first problem is that Digikam can't make good suggestions > 2- There is not a visual differentiation between faces already confirmed and > faces suggested by digikam and not confirmed yet. Because of it you need to > move the cursor over the picture to figure out if it is already confirmed > or not. > > Nobody else think that this is a problem? > > Cheers > Erick Moreno > > On Fri, May 20, 2011 at 6:23 AM, sleepless > > Hi Erick, > > > > In my experience digikam always suggest the latest applied face tag for > > the next one until you applie another one from then on it will suggest > > that tag. So face detection is great, recognition not at this point in > > my opinion (I would not recommend 100% setting but rather default). > > > > It should be nice to hear how it went to other people as well. > > > > Not very helpfull but it may be not you who is to blame. > > > > Rinus > > > > > > > > Op 20-05-11 03:07, Erick Moreno schreef: > > > > Hi, > > > > I think that I'm doing something wrong with face detection using Digikam > > 2.0.0-beta5. > > > > Just after update to the new beta version, I tried to recognize some > > faces in my photos. I made a first scan to "Detect and recognize > > faces". This scan takes about 12hs. After that, I marked some people > > (about 10), including me, and tried to recognize faces. > > > > The result was quite disappointing, digikam suggested to mark with my > > tag, a lot of faces from people that really doesn't look like me. I'm a > > 28years old guy from Brazil and it suggested 80y old ladies, dogs, some > > japanese friends, children, and go on. > > > > At first I thought that the algorithm needed more information to be > > accurate and I confirmed a lot of pictures of me (about 200) and created > > others tags to mark people that it confuses with me with no success. > > > > I made a lot of scans, merged results, tried to recognize faces again, > > set the accuracy to 100, cleared unconfirmed results and rescan, > > confirmed about 500 photos manually and I still can't see the face > > recognition match faces correctly. > > > > What I'm doing wrong? > > > > Cheers > > > > > > -- > > *Erick Moreno* > > https://profiles.google.com/erickmoreno > > > > > > _______________________________________________ > > Digikam-users mailing > > [hidden email]://mail.kde.org/mailman/listinfo/digikam- > > users > > > > > > > > _______________________________________________ > > Digikam-users mailing list > > [hidden email] > > https://mail.kde.org/mailman/listinfo/digikam-users _______________________________________________ Digikam-users mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-users |
Re: What I'm doing wrong with face detection?
|
|
In reply to this post by Erick Moreno
Am Montag, 23. Mai 2011, 12:00:32 schrieb Erick Moreno:
> I noticed that too, this happens when you are tagging faces. But, when you > try to match new faces with existent tags, the face detection suggest new > faces for that tag and put the suggestions together with the faces already > confirmed. I see some problems in the way that Digikam is doing this. > > 1- Of course the first problem is that Digikam can't make good suggestions > 2- There is not a visual differentiation between faces already confirmed and > faces suggested by digikam and not confirmed yet. Because of it you need to > move the cursor over the picture to figure out if it is already confirmed > or not. > > Nobody else think that this is a problem? I think that there are a lot of issues with current implementation of face detection, user guide to the first scan and the face detection features, the time it takes to get first results, no grouping to ease naming etc. One simply has to use Google's Picasa once to see how easy to use and unhidden it should be in the GUI. But digikam is different in terms of resources etc. thus the first version of face detection will have issues when it is released, that's the way open- source works, incremental improvements over the releases. Sven PS: Sorry for the other reply, I clicked on send by mistake. _______________________________________________ Digikam-users mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-users |
«
Return to digikam-users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |