Menu
▾
▴
rtftr-devel — Discussions for rtftr developers
You can subscribe to this list here.
| 2008 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(9) |
Oct
(26) |
Nov
(14) |
Dec
(79) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2009 |
Jan
(6) |
Feb
(7) |
Mar
(55) |
Apr
(13) |
May
|
Jun
|
Jul
|
Aug
(1) |
Sep
|
Oct
|
Nov
|
Dec
|
| 2010 |
Jan
|
Feb
|
Mar
(2) |
Apr
(2) |
May
|
Jun
(1) |
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2011 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
(2) |
Sep
|
Oct
|
Nov
|
Dec
|
| 2012 |
Jan
|
Feb
|
Mar
(7) |
Apr
(2) |
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: lakesh <lak...@gm...> - 2008-12-12 09:18:05
|
Hi, This MYRA people have also tried to implement a real time face tracking and recognition using the algorithms that we are currently implementing. Have a look at this http://74.125.77.132/search?q=cache:3e8j_q67YLYJ:www.bmf.hu/conferences/sisy2006/17_Vamossy.pdf+similarity+measures+using+Gabor+filters+face+recognition&hl=ne&ct=clnk&cd=24&gl=np -- Have a nice day |
|
From: lakesh <lak...@gm...> - 2008-12-10 13:34:24
|
http://www.it.lut.fi/project/simplegabor/downloads/src/simplegabortb/ This software can be helpful for analyzing Gabor filters. Check this out -- Have a nice day |
|
From: Anjan <roc...@gm...> - 2008-12-09 15:35:36
|
Hi Bibek, I have extracted the images from the .vec files. The two databases have the size of 3.5 MB each after extraction. If you need it let me know. -- Anjan Nepal GPO Box 11458 Registered Linux User #395963 |
|
From: Abhishek D. <the...@gm...> - 2008-12-09 09:59:03
|
Hi lakesh, I wish if you could generate a plot that shows the distance of probe image with respect to all the training face images. Other plots that will help in visualizing the performance will also be helpful. Put the plot in wiki after you finish testing. bye On Tue, Dec 9, 2008 at 12:56 PM, Abhishek Dutta <the...@gm...>wrote: > > > On Tue, Dec 9, 2008 at 12:31 PM, Abhishek Dutta <the...@gm...>wrote: > >> Ok wait >> >> >> >> On Tue, Dec 9, 2008 at 12:26 PM, lakesh <lak...@gm...>wrote: >> >>> Hi Abhishek, >>> Can you also upload the plots like the one before so that it is easier >>> for me to check if the results are on the right track or not >>> >>> >>> On Tue, Dec 9, 2008 at 12:24 PM, Abhishek Dutta < >>> the...@gm...> wrote: >>> >>>> >>>> >>>> On Tue, Dec 9, 2008 at 9:53 AM, lakesh <lak...@gm...>wrote: >>>> >>>>> Ok Abhishek, >>>>> I will try to get the results by today and inform u. >>>>> >>>>> On Tue, Dec 9, 2008 at 9:50 AM, Abhishek Dutta < >>>>> the...@gm...> wrote: >>>>> >>>>>> Hi lakesh, >>>>>> The training and probe images are located at >>>>>> >>>>>> https://apps.sourceforge.net/mediawiki/rtftr/index.php?title=EBGM_Implementation >>>>>> >>>>>> bye >>>>>> >>>>>> -- >>>>>> Abhishek >>>>>> http://adutta.np.googlepages.com >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------------ >>>>>> SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, >>>>>> Nevada. >>>>>> The future of the web can't happen without you. Join us at MIX09 to >>>>>> help >>>>>> pave the way to the Next Web now. Learn more and register at >>>>>> >>>>>> http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ >>>>>> _______________________________________________ >>>>>> Rtftr-devel mailing list >>>>>> Rtf...@li... >>>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Have a nice day >>>>> >>>> >>>> >>>> >>>> -- >>>> Abhishek >>>> http://adutta.np.googlepages.com >>>> >>> >>> >>> >>> -- >>> Have a nice day >>> >> >> >> >> -- >> Abhishek >> http://adutta.np.googlepages.com >> > > > > -- > Abhishek > http://adutta.np.googlepages.com > -- Abhishek http://adutta.np.googlepages.com |
|
From: Abhishek D. <the...@gm...> - 2008-12-09 04:06:01
|
Hi lakesh, The training and probe images are located at https://apps.sourceforge.net/mediawiki/rtftr/index.php?title=EBGM_Implementation bye -- Abhishek http://adutta.np.googlepages.com |
|
From: Abhishek D. <adu...@gm...> - 2008-12-08 00:24:58
|
We'll meet tomorrow after IS lab time. bye On Sun, Dec 7, 2008 at 10:05 PM, Anjan <roc...@gm...> wrote: > Hi all, > I have completed my code on adaboost based face detection in matlab > according to the paper on Viola and Jones. I have not been able to get the > features as explained in the paper. So, Bibek and I will be working in this > algorithm together from tomorrow, leaving NN for later. This might change > our schedule (especially lakesh's). So, should we have a brief meeting > tomorrow on discussing our schedule, or the project plan can be discussed > through the mailing list only? Reply soon. > > -- > Anjan Nepal > > GPO Box 11458 > Registered Linux User #395963 > > > > ------------------------------------------------------------------------------ > SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada. > The future of the web can't happen without you. Join us at MIX09 to help > pave the way to the Next Web now. Learn more and register at > > http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Abhishek http://adutta.np.googlepages.com |
|
From: Anjan <roc...@gm...> - 2008-12-07 16:20:47
|
Hi all, I have completed my code on adaboost based face detection in matlab according to the paper on Viola and Jones. I have not been able to get the features as explained in the paper. So, Bibek and I will be working in this algorithm together from tomorrow, leaving NN for later. This might change our schedule (especially lakesh's). So, should we have a brief meeting tomorrow on discussing our schedule, or the project plan can be discussed through the mailing list only? Reply soon. -- Anjan Nepal GPO Box 11458 Registered Linux User #395963 |
|
From: Abhishek D. <the...@gm...> - 2008-12-07 15:41:13
|
Hi all, i have uploaded the new version of ioe face database in our website. Check details at https://apps.sourceforge.net/mediawiki/rtftr/index.php?title=Meeting_Dec_03_2008 bye === contents of readme.txt ====== ############################ IOE Face Database ################################# This database has been created for the testing purpose of the final year project "Real Time Face Tracking & Recognition (rtftr)" ( http://rtftr.sourceforge.net ) Photographs taken on : 12:00 midday , 3 December 2008 Place : IOE Library Number of subjects : 13 Last Updated : 7 December 2008 Total no. of images : 227 # Image Naming convention # f0010v009 | |________> pose id (can vary from 001 to 999) |_____________> face id (can vary from 0001 to 9999) # Description of contents of Face Database . |-- images --------> training face images | |-- f0001v001.jpg | |-- f0001v002.jpg | |-- ...... | |-- f0013v016.jpg | `-- f0013v017.jpg |-- probe_images --------> group photos that can be used for testing (probe) | |-- dsc04448.jpg | |-- dsc04450.jpg | `-- dsc04451.jpg `-- readme.txt --------> this file # Details of subjects in the face database +==============================================================================+ | Face Id | Name | Class Roll No | No. of pose variations | +==============================================================================+ | 0001 | Niraj Sapkota | 062BCT521 | 015 | | 0002 | Sushil Shilpakar | 062BCT547 | 023 | | 0003 | Sabbir Kumar Manandhar| 062BCT537 | 013 | | 0004 | Saroj Gautam | 062BCT541 | 014 | | 0005 | Bibek Raj Dhakal | 062BCT506 | 016 | | 0006 | Subash Basnet | 062BCT544 | 018 | | 0007 | Madhusudan Ram Joshi | 062BCT518 | 020 | | 0008 | Ruchin Singh | 062BCT536 | 020 | | 0009 | Bibek Karmacharya | 061BCT508 | 016 | | 0010 | Bibek Shrestha | 061BCT509 | 021 | | 0011 | Lakesh Kansakar | 061BCT521 | 017 | | 0012 | Anjan Nepal | 061BCT502 | 017 | | 0013 | Abhishek Dutta | 061BCT501 | 017 | +==============================================================================+ -- Abhishek http://adutta.np.googlepages.com |
|
From: Anjan <roc...@gm...> - 2008-12-05 05:56:09
|
Hey, Let's put the IOE face database online. I think ioelive.com will be the suitable place. We can scale down the pictures and crop if necessary before uploading. -- Anjan Nepal GPO Box 11458 Registered Linux User #395963 |
|
From: Anjan <roc...@gm...> - 2008-12-05 04:56:29
|
Hey guys, I have completed my cascade detector code in matlab. cascade detector is the collection of several classifiers which are arranged in several stages. The classifiers are selected using adaboost algorithm. Currently, i'm training the cascade detector, but it takes a really long time. In minimum data too, it takes around 2-3 hours for the complete training. And, even more, the results are not promising. I think if more data are added the performance will improve (not in the time, but detection). I'm working on how can i improve the speed too. Anyone has cropped the faces that we collected in the library? If anyone has done that already, i'd like to include them in my training set. -- Anjan Nepal GPO Box 11458 Registered Linux User #395963 |
|
From: Abhishek D. <the...@gm...> - 2008-12-05 03:01:49
|
Hi all, see some results at this wiki page http://apps.sourceforge.net/mediawiki/rtftr/index.php?title=EBGM_Implementation The results are not very promising at this moment. But I am sure it will be after some fine tuning. bye -- Abhishek http://adutta.np.googlepages.com |
|
From: lakesh <lak...@gm...> - 2008-12-04 15:39:52
|
Also Bibek if you can provide me the detected faces as well as non faces soon ,then I can start the thresholding stuff so that it is easier later. Ok guys, keep the mailing list busy :) On Thu, Dec 4, 2008 at 9:21 PM, lakesh <lak...@gm...> wrote: > Hi all, > I was searching for libraries to use for GUI in our project. One option was > wxwidget. Another is FLTK, it is light weight. But it provides the basics > that we want. Here is a link for an face recognition app using FLTK and > opencv. Check this out. > http://derin-deli-mavi.googlecode.com/files/FaceRec.tar.gz > The GUI is very simple. Wxwidget provides advanced gui stuffs but it takes > control over everything. We cannot even define a main function when using > wxwidget. It has it's own macros and class that we need to initialize. It > holds the main function itself. > Here is some comparison > > http://samibadawi.blogspot.com/2008/11/computer-vision-c-libraries-review.html > > -- > Have a nice day > -- Have a nice day |
|
From: lakesh <lak...@gm...> - 2008-12-04 15:36:59
|
Hi all, I was searching for libraries to use for GUI in our project. One option was wxwidget. Another is FLTK, it is light weight. But it provides the basics that we want. Here is a link for an face recognition app using FLTK and opencv. Check this out. http://derin-deli-mavi.googlecode.com/files/FaceRec.tar.gz The GUI is very simple. Wxwidget provides advanced gui stuffs but it takes control over everything. We cannot even define a main function when using wxwidget. It has it's own macros and class that we need to initialize. It holds the main function itself. Here is some comparison http://samibadawi.blogspot.com/2008/11/computer-vision-c-libraries-review.html -- Have a nice day |
|
From: Abhishek D. <adu...@gm...> - 2008-12-04 13:50:32
|
Hi Marc, We discussed about getting involved in the "Face recognition for Facebook" project. However at present we are very busy with "rtftr<http://rtftr.sourceforge.net>" project. As this is a final year project for our Bachelor in Computer Engineering degree, we have to complete it in time (everything should be ready by Jan15, 2009). We can only participate in the project after finishing "rtftr <http://rtftr.sourceforge.net/>" project. However, we will be in touch with the ongoing development in the project via google groups <http://groups.google.com/group/fr4fb?hl=en>. Bye Abhishek On Sat, Nov 29, 2008 at 11:48 PM, Marc . <mar...@gm...> wrote: > Hi Abhishek, > Just giving you an update about the current "Face recognition for > Facebook" project status. > > There are tow people that already agreed to participate in the project > (get a look in the mails below) and the "official" web page of the > project will be Faint ( https://sourceforge.net/projects/faint/ ) > where I already open a discussion in the forum there. > > So, please sign on there ( or just reply for my last post in the > forum) if you wish to get involved (it's a good opportunity to get a > perfect final year project by the way ; ) > > Ok, thanks for your attention and any doubt, please contact me. > Hope to hear from you! > > > > Best Regards, > Marc Rosenfeld > > PS: I hide the mails below in order to keep their privacy.. > > > >>>>>>>>>>>>>>>>>>> > > > > Ok, > > Apparently everybody is in: Malte, Gino, Mino and Marc (me:) and the > "official" webpage is Faint http://sourceforge.net/projects/faint (I > will post a topic for the "Faint4FB" project there). > > Gino, I studied management and, according to the "knowledge sharing > principles", I must say that we will probably get a better results if > the project is pure GPL/ open source, why? well, if we make a > commercial product, it will be hard to have more contributors, and > really, when talking about "pure knowledge" (like algorithms) a better > result is reached if we get a maximum number of people with different > ideas to contributing for the same purpose (here, making a good face > recognition system ;). > > Malte, I tested tag4you: actually the app only detects faces, it > doesn't recognize them. Even if they are looking to make a face rec. > system, this function is still in beta and available only under > request (I ask for it a few days ago but I still not get any answer > from them) so really, I think a "Faint 4 FB" would start at least as > good as tag4you, but will be reach a much better result later: we are > able to get many support for the purpose here, they aren't! ; ) > > Ok, that's all for now, we keep in touch in Faint's page! > > Best Regards, > Marc Rosenfeld > > PS: I am in Brazil ( and yes it's sunny and warm here ; ) > - Masquer le texte des messages précédents - > > > >>>>> > > Hi Gino and Marc! > > I really like the idea of a collaboration between fb4j and faint and > it's good to see everybody being enthusiastic about it :) Maybe we > should just start sketching something like a simple prototype of that > facebook app. > > As Marc already mentioned I do not have any experience in creating > applications for Facebook yet but I'm sure some modifications of the > faint library are necessary since it was initially designed to work > with images on a local filesystem. Another problem will probably be > the current OpenCV plugin for face detection which depends on native C > code and might cause trouble on some operating systems, especially if > the detection is done on client side. Is it even possible to run code > on client side in Facebook? > > A good idea is probably also to prevent reinventing the wheel... There > is already some kind of face detection software for Facebook, it's > called "Tag4You". I did not have the chance to test this one yet, but > if the results for detection and recognition are better than what you > can do with faint there is no need for the integration at all. > > Well, both of you are welcome to join the faint 'team' on SourceForge > of course so we could use the svn repository, forums and stuff like > that. I'm currently living in northern Germany, near Bremen so the > chance for a developer meeting is rather small I guess. > > Best regards, > Malte > > @Marc: I'll contact you later about the errors you reported. By the > way: Although 'Malte' is also a name for girls I'm actually not > female! But don't worry about that misunderstanding. :) > > > > 2008/11/29 Gino Miceli <>: > - Masquer le texte des messages précédents - > > Hey Marc, > > Of course, please send away! Use this one, my Gmail addy. > > My motivations for wanting to create a face detection tool for Facebook > were > > identical to yours. I even implemented a highly accurate face detection > > algorithm using Support Vector Machines, but like I said, never got > around > > to creating a Facebook app. > > fb4j 2 is still under construction, but if we work together we can be > sure > > to implement the parts needed for the project first. > > I'm CC'ing my partner in fb4j on this email.. he's the other half of the > > team and knows a lot about fb4j 1 and 2. > > What is your background? Do you have a business plan or any ideas of > > whether this would be commercial or free, licenses, etc? > > Keep in touch, > > -Gino > > > > > > > > On Fri, Nov 28, 2008 at 8:20 PM, Marc . <> wrote: > >> > >> Hi Gino, > >> > >> First of all, I must say that I am "only" a big fan of Faint (as I > >> told on the tread ) however I am not a developer, just a enthusiast > >> (sorry if I disappointed you here ; ). > >> > >> In fact this project came in my mind when I was discussing with Malte > >> (the Faint's developer) about customizing Faint to accept command > >> lines. During this project I proposed Malte to adapt Faint in order to > >> create a Facebook App, she answered me that this would be a great > >> idea, but she claims that she doesn't have much experience about > >> creating FB apps, that why I came to you (Gino). > >> > >> Ok, that all for the quick history, but I think that if both of you > >> agree (Gino, Malte is in Bcc copy of this mail, Malte, Gino is in Bcc > >> copy of this mail), we could create a good face recognition FB app. > >> > >> In fact I think that FB is "The" ideal environment to create such a > >> face recognition system because it has 3 important characteristics: > >> > >> 1- Tons of photos with faces, took in a variety of ambiances with > >> different angles and luminosity (ideal to test face recognition > >> algorithms). > >> > >> 2- Tons of labels already done "by hand" by the users, that could > >> easily constitute an important initial "face database" in order to > >> "teach" the software to recognize friends. > >> > >> 3- Tons of users and enthusiasts that would love to contribute in such > >> a face recognition application (I just found 2 of them in less than 24 > >> hours! ;) > >> > >> Now to the questions: Gino, can I send your mail address to Malte? > >> Malte, can I send your address to Gino? > >> By the way, I received a couple of other supports for this project > >> today, please let me know if we could use one of the 2 existing > >> projects (FB4J or Faint) as the "official" meeting place. > >> > >> > >> Hope to hear from you! > >> > >> Best Regards, > >> Marc Rosenfeld > >> > >> > >> 2008/11/28 Gino Miceli <>: > >> > Brilliant! This is exactly the app I wanted to create! I just got my > >> > specialization in Machine Learning and have learned some face > >> > recognition > >> > and detection techniques, but haven't had time to implement anything > >> > yet.. > >> > need any help with your project? Where are you based? > >> > FYI, we're working on fb4j 2.0 right now. The new website will be at > >> > www.fb4j.org. Stay tuned. > >> > -Gino > >> > > >> > On Fri, Nov 28, 2008 at 5:02 AM, Marc R > >> > <> wrote: > >> >> > >> >> Message body follows: > >> >> > >> >> Hi Gino, > >> >> > >> >> I found you in the fb4j project ( > >> >> http://fb4j.sourceforge.net/ ) here in Source Forge and I > >> >> think you might be able to help me in a new project: adapt a > >> >> Face Recognition app. to work in Facebook (the Faint, it's a > >> >> open source software written in Java, thus your API just fit > >> >> the needs ; ) > >> >> > >> >> I just open a new forum thread about the idea, get a look > >> >> and tell me what you think (any opinion is a help : ) > >> >> > >> >> > https://sourceforge.net/community/forum/topic.php?id=4195&page&replies=1 > >> >> > >> >> Anyway thanks for your attention and any doubt, please > >> >> contact me. > >> >> > >> >> Best Regards, > >> >> Marc Rosenfeld > >> >> > >> >> PS: If you wish, you can reply me by mail at mar...@gm... > >> >> > >> >> -- > >> >> This message has been sent to you, a registered SourceForge.net user, > >> >> by another site user, through the SourceForge.net site. This message > >> >> has been delivered to your SourceForge.net mail alias. You may reply > >> >> to this message using the "Reply" feature of your email client, or > >> >> using the messaging facility of SourceForge.net at: > >> >> > >> >> > >> > > >> > > > > > > > > > 2008/11/28 Marc . <mar...@gm...>: > > Hi Dutta, > > Ok, let me know when you discussed with your project members and > > please, keep in touch for any doubt or question. > > By the way, I sent them a message yesterday about this subject, so I > > think they may be already aware about it. > > Hope to hear from you. > > > > Best Regards, > > Marc Rosenfeld > > > > > > 2008/11/28 Abhishek Dutta <adu...@gm...>: > >> Hi Marc, > >> The Facebook application idea is very interesting. Let me first discuss > with > >> my project members and then I will inform you about what we think. At > >> present we are a bit busy with our final year project. > >> > >> Bye > >> > >> On Fri, Nov 28, 2008 at 9:36 AM, Marc R > >> <ros...@us...> wrote: > >>> > >>> Message body follows: > >>> > >>> Hi Dutta, > >>> > >>> I found you in the RTFTR project ( > >>> http://rtftr.sourceforge.net/ ) and I thought you might be > >>> interested in participate in my project idea as there is a > >>> number or synergies (and similarities) with your project: > >>> create a Face Recognition application for Facebook. > >>> > >>> The main idea is to create a auto-labeling app. that could > >>> indicated friend's name in a photo. > >>> > >>> Get a look at > >>> > https://sourceforge.net/community/forum/topic.php?id=4195&page&replies=1 > >>> and tell me what you think. > >>> Really, I think an app like this can be very popular among > >>> Faceboook users. > >>> > >>> Anyway, thanks for your attention and any doubt, please > >>> contact me. > >>> > >>> Best Regards, > >>> Marc Rosenfeld > >>> > >>> PS: if you wish, you can reply me by mail at mar...@gm... > >>> > >>> -- > >>> This message has been sent to you, a registered SourceForge.net user, > >>> by another site user, through the SourceForge.net site. This message > >>> has been delivered to your SourceForge.net mail alias. You may reply > >>> to this message using the "Reply" feature of your email client, or > >>> using the messaging facility of SourceForge.net at: > >>> https://sourceforge.net/sendmessage.php?touser=2263069 > >>> > >> > >> > >> > >> -- > >> Abhishek > >> http://adutta.np.googlepages.com > >> > > > -- Abhishek http://adutta.np.googlepages.com |
|
From: Abhishek D. <adu...@gm...> - 2008-12-04 01:52:07
|
Hi all, the location we photographed our subjects is good for working. It has power and a good place to work. So we all can sit there and coordinate. I am coming to library at 11:00AM. bye On Wed, Dec 3, 2008 at 10:36 PM, Bibek Shrestha <bib...@gm...>wrote: > thanks for the link lakesh, > > im running the NN over the images, i'll update you on the results. > > cheers > > On Wednesday 03 December 2008 18:40:45 lakesh wrote: > > > > http://jwalanta.blogspot.com/2007/03/batch-jpeg-image-resize-in-linux-jpeg > . > > >html This script from Jwalanta's blog is useful to scale the images. Lets > > > scale the images and apply the NN detector. > > > Can it be done by tomorrow?? > > > > > > On Wed, Dec 3, 2008 at 6:38 PM, lakesh <lak...@gm...> > wrote: > > > > Yes, Bibek we should do it soon. The images from Abhishek's cam, they > are > > > > of very high resolution. We can scale them down so that it is easier > for > > > > the NN detector to detect. Some of the subjects they would be very > tough > > > > to detect because of their positions. The others they should be > > > > detectable I guess. > > > > > > > > On Wed, Dec 3, 2008 at 6:27 PM, Abhishek Dutta <adu...@gm... > >wrote: > > > >> Hi Lakesh, > > > >> Coordinate with bibek and perform the testing part. I am busy with the > > > >> matlab implementation and sudy of Gabor wavelet. We all can meet if it > > > >> necessary. > > > >> > > > >> bye > > > >> > > > >> On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...> > wrote: > > > >>> Hi all, > > > >>> I went through the images. Now what we need to do is the NN detector > > > >>> will detect the images and pass it to the subspace LDA for training. > > > >>> Lets do it as soon as possible. I found that the ORL database is also > > > >>> not perfectly registered. There are slight rotations in the images of > a > > > >>> person. However in our case we have lots of rotations. Some of the > > > >>> images they may not be useful because I don't think that NN detector > > > >>> will detect them. However the frontal faces they should be detected. > > > >>> Lets do this test. > > > >>> > > > >>> -- > > > >>> Have a nice day > > > >>> > > > >>> > ----------------------------------------------------------------------- > > > >>>-- This SF.Net email is sponsored by the Moblin Your Move Developer's > > > >>> challenge > > > >>> Build the coolest Linux based applications with Moblin SDK & win > great > > > >>> prizes > > > >>> Grand prize is a trip for two to an Open Source event anywhere in the > > > >>> world > > > >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ > > > >>> _______________________________________________ > > > >>> Rtftr-devel mailing list > > > >>> Rtf...@li... > > > >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > > >> > > > >> -- > > > >> Abhishek > > > >> http://adutta.np.googlepages.com > > > > > > > > -- > > > > Have a nice day > > -- > > Bibek Shrestha > > bib...@gm... > > 9841 825 925 > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Abhishek http://adutta.np.googlepages.com |
|
From: Bibek S. <bib...@gm...> - 2008-12-03 16:51:59
|
thanks for the link lakesh, im running the NN over the images, i'll update you on the results. cheers On Wednesday 03 December 2008 18:40:45 lakesh wrote: > http://jwalanta.blogspot.com/2007/03/batch-jpeg-image-resize-in-linux-jpeg. >html This script from Jwalanta's blog is useful to scale the images. Lets > scale the images and apply the NN detector. > Can it be done by tomorrow?? > > On Wed, Dec 3, 2008 at 6:38 PM, lakesh <lak...@gm...> wrote: > > Yes, Bibek we should do it soon. The images from Abhishek's cam, they are > > of very high resolution. We can scale them down so that it is easier for > > the NN detector to detect. Some of the subjects they would be very tough > > to detect because of their positions. The others they should be > > detectable I guess. > > > > On Wed, Dec 3, 2008 at 6:27 PM, Abhishek Dutta <adu...@gm...>wrote: > >> Hi Lakesh, > >> Coordinate with bibek and perform the testing part. I am busy with the > >> matlab implementation and sudy of Gabor wavelet. We all can meet if it > >> necessary. > >> > >> bye > >> > >> On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...> wrote: > >>> Hi all, > >>> I went through the images. Now what we need to do is the NN detector > >>> will detect the images and pass it to the subspace LDA for training. > >>> Lets do it as soon as possible. I found that the ORL database is also > >>> not perfectly registered. There are slight rotations in the images of a > >>> person. However in our case we have lots of rotations. Some of the > >>> images they may not be useful because I don't think that NN detector > >>> will detect them. However the frontal faces they should be detected. > >>> Lets do this test. > >>> > >>> -- > >>> Have a nice day > >>> > >>> ----------------------------------------------------------------------- > >>>-- This SF.Net email is sponsored by the Moblin Your Move Developer's > >>> challenge > >>> Build the coolest Linux based applications with Moblin SDK & win great > >>> prizes > >>> Grand prize is a trip for two to an Open Source event anywhere in the > >>> world > >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ > >>> _______________________________________________ > >>> Rtftr-devel mailing list > >>> Rtf...@li... > >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel > >> > >> -- > >> Abhishek > >> http://adutta.np.googlepages.com > > > > -- > > Have a nice day -- Bibek Shrestha bib...@gm... 9841 825 925 |
|
From: lakesh <lak...@gm...> - 2008-12-03 13:02:45
|
Check the ORL database. FR is giving good results for that database. I don't think they are registered according to the eye position. However there are only slight out of plane rotation. So lets give it a try On Wed, Dec 3, 2008 at 6:40 PM, lakesh <lak...@gm...> wrote: > > http://jwalanta.blogspot.com/2007/03/batch-jpeg-image-resize-in-linux-jpeg.html > This script from Jwalanta's blog is useful to scale the images. Lets scale > the images and apply the NN detector. > Can it be done by tomorrow?? > > > On Wed, Dec 3, 2008 at 6:38 PM, lakesh <lak...@gm...> wrote: > >> Yes, Bibek we should do it soon. The images from Abhishek's cam, they are >> of very high resolution. We can scale them down so that it is easier for the >> NN detector to detect. Some of the subjects they would be very tough to >> detect because of their positions. The others they should be detectable I >> guess. >> >> >> On Wed, Dec 3, 2008 at 6:27 PM, Abhishek Dutta <adu...@gm...>wrote: >> >>> Hi Lakesh, >>> Coordinate with bibek and perform the testing part. I am busy with the >>> matlab implementation and sudy of Gabor wavelet. We all can meet if it >>> necessary. >>> >>> bye >>> >>> On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...>wrote: >>> >>>> Hi all, >>>> I went through the images. Now what we need to do is the NN detector >>>> will detect the images and pass it to the subspace LDA for training. Lets do >>>> it as soon as possible. I found that the ORL database is also not perfectly >>>> registered. There are slight rotations in the images of a person. However in >>>> our case we have lots of rotations. Some of the images they may not be >>>> useful because I don't think that NN detector will detect them. However the >>>> frontal faces they should be detected. Lets do this test. >>>> >>>> -- >>>> Have a nice day >>>> >>>> >>>> ------------------------------------------------------------------------- >>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>> challenge >>>> Build the coolest Linux based applications with Moblin SDK & win great >>>> prizes >>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>> world >>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>> _______________________________________________ >>>> Rtftr-devel mailing list >>>> Rtf...@li... >>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>> >>>> >>> >>> >>> -- >>> Abhishek >>> http://adutta.np.googlepages.com >>> >> >> >> >> -- >> Have a nice day >> > > > > -- > Have a nice day > -- Have a nice day |
|
From: lakesh <lak...@gm...> - 2008-12-03 12:55:48
|
http://jwalanta.blogspot.com/2007/03/batch-jpeg-image-resize-in-linux-jpeg.html This script from Jwalanta's blog is useful to scale the images. Lets scale the images and apply the NN detector. Can it be done by tomorrow?? On Wed, Dec 3, 2008 at 6:38 PM, lakesh <lak...@gm...> wrote: > Yes, Bibek we should do it soon. The images from Abhishek's cam, they are > of very high resolution. We can scale them down so that it is easier for the > NN detector to detect. Some of the subjects they would be very tough to > detect because of their positions. The others they should be detectable I > guess. > > > On Wed, Dec 3, 2008 at 6:27 PM, Abhishek Dutta <adu...@gm...>wrote: > >> Hi Lakesh, >> Coordinate with bibek and perform the testing part. I am busy with the >> matlab implementation and sudy of Gabor wavelet. We all can meet if it >> necessary. >> >> bye >> >> On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...> wrote: >> >>> Hi all, >>> I went through the images. Now what we need to do is the NN detector will >>> detect the images and pass it to the subspace LDA for training. Lets do it >>> as soon as possible. I found that the ORL database is also not perfectly >>> registered. There are slight rotations in the images of a person. However in >>> our case we have lots of rotations. Some of the images they may not be >>> useful because I don't think that NN detector will detect them. However the >>> frontal faces they should be detected. Lets do this test. >>> >>> -- >>> Have a nice day >>> >>> ------------------------------------------------------------------------- >>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>> challenge >>> Build the coolest Linux based applications with Moblin SDK & win great >>> prizes >>> Grand prize is a trip for two to an Open Source event anywhere in the >>> world >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>> _______________________________________________ >>> Rtftr-devel mailing list >>> Rtf...@li... >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>> >>> >> >> >> -- >> Abhishek >> http://adutta.np.googlepages.com >> > > > > -- > Have a nice day > -- Have a nice day |
|
From: lakesh <lak...@gm...> - 2008-12-03 12:53:49
|
Yes, Bibek we should do it soon. The images from Abhishek's cam, they are of very high resolution. We can scale them down so that it is easier for the NN detector to detect. Some of the subjects they would be very tough to detect because of their positions. The others they should be detectable I guess. On Wed, Dec 3, 2008 at 6:27 PM, Abhishek Dutta <adu...@gm...> wrote: > Hi Lakesh, > Coordinate with bibek and perform the testing part. I am busy with the > matlab implementation and sudy of Gabor wavelet. We all can meet if it > necessary. > > bye > > On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...> wrote: > >> Hi all, >> I went through the images. Now what we need to do is the NN detector will >> detect the images and pass it to the subspace LDA for training. Lets do it >> as soon as possible. I found that the ORL database is also not perfectly >> registered. There are slight rotations in the images of a person. However in >> our case we have lots of rotations. Some of the images they may not be >> useful because I don't think that NN detector will detect them. However the >> frontal faces they should be detected. Lets do this test. >> >> -- >> Have a nice day >> >> ------------------------------------------------------------------------- >> This SF.Net email is sponsored by the Moblin Your Move Developer's >> challenge >> Build the coolest Linux based applications with Moblin SDK & win great >> prizes >> Grand prize is a trip for two to an Open Source event anywhere in the >> world >> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >> _______________________________________________ >> Rtftr-devel mailing list >> Rtf...@li... >> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >> >> > > > -- > Abhishek > http://adutta.np.googlepages.com > -- Have a nice day |
|
From: Abhishek D. <adu...@gm...> - 2008-12-03 12:42:43
|
Hi Lakesh, Coordinate with bibek and perform the testing part. I am busy with the matlab implementation and sudy of Gabor wavelet. We all can meet if it necessary. bye On Wed, Dec 3, 2008 at 6:21 PM, lakesh <lak...@gm...> wrote: > Hi all, > I went through the images. Now what we need to do is the NN detector will > detect the images and pass it to the subspace LDA for training. Lets do it > as soon as possible. I found that the ORL database is also not perfectly > registered. There are slight rotations in the images of a person. However in > our case we have lots of rotations. Some of the images they may not be > useful because I don't think that NN detector will detect them. However the > frontal faces they should be detected. Lets do this test. > > -- > Have a nice day > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Abhishek http://adutta.np.googlepages.com |
|
From: lakesh <lak...@gm...> - 2008-12-03 12:36:28
|
Hi all, I went through the images. Now what we need to do is the NN detector will detect the images and pass it to the subspace LDA for training. Lets do it as soon as possible. I found that the ORL database is also not perfectly registered. There are slight rotations in the images of a person. However in our case we have lots of rotations. Some of the images they may not be useful because I don't think that NN detector will detect them. However the frontal faces they should be detected. Lets do this test. -- Have a nice day |
|
From: Abhishek D. <adu...@gm...> - 2008-12-03 02:01:35
|
Hi all, I talked with Bibek and he is also ready for today's photographing task. Let's meet at 10:30 AM today in college. bye On Wed, Dec 3, 2008 at 7:07 AM, lakesh <lak...@gm...> wrote: > +1 > > > On Tue, Dec 2, 2008 at 9:32 PM, Anjan <roc...@gm...> wrote: > >> ACK >> >> >> On Tue, Dec 2, 2008 at 9:07 PM, Abhishek Dutta <adu...@gm...>wrote: >> >>> Ok guys >>> I am ready for clicking photos tomorrow. Moreover, I will also bring my >>> camcorder so that we can have a video of people being photographed. We can >>> use this video to do testing in later part. >>> >>> Everybody ACK if you are ready. >>> >>> bye >>> >>> >>> On Tue, Dec 2, 2008 at 8:42 PM, lakesh <lak...@gm...>wrote: >>> >>>> Hi all, >>>> What I actually want to test are the following. >>>> 1) Most of the people, when they build a face detection and recognition >>>> system use some threshold value to separate the in database and not in >>>> database classification. This threshold directly depends on the training >>>> sets. So for different sets different thresholds. So obviously there is >>>> double work if we go for different data sets later. >>>> 2) We haven't done geometric normalization which most of the people do. >>>> Based on the paper Anjan referred we are going to take some detected images >>>> of a person to train SubspaceLda. We don't know how this will perform. So I >>>> want to check this as well >>>> 3) I just want to make sure that we know where we are heading and be >>>> prepared. We are using different research papers to implement the >>>> algorithms. Specifically in the case of fr most of the papers they aren't >>>> considering face detection part. They have just tested in the standard >>>> databases. But in our case we are building an actual system where we have to >>>> consider lots of things. JT sir said that it is ok even if we have 50% >>>> accuracy. So it's a word of relief for us. However before going for video if >>>> we can just test using the still images we will get some hints I guess. >>>> >>>> >>>> On Tue, Dec 2, 2008 at 8:25 PM, Abhishek Dutta <adu...@gm...>wrote: >>>> >>>>> Hi all, >>>>> I agree to all Anjan's points. Phase 1 integration is necessary. And >>>>> lakesh seems to be waiting for bibek. >>>>> >>>>> Moreover Lakesh want to photograph subjects right now to find the >>>>> "threshold" values (I guess). So 10 subjects with 10 pose variation means >>>>> total 100 photos. This is not a huge task. I can bring my camera and bibek >>>>> will bring his stand. We can then photograph 10 people. Now we have to >>>>> decide the following stuffs >>>>> >>>>> # which 10 people: I am also not in favor of photographing our class >>>>> mates as they might be busy during final testing phase (Jan 01, 2009 to Jan >>>>> 15, 2009). So we have to find volunteers. Bibek has lots of contacts from >>>>> 3rd year. He can arrange that. We have to keep in mind that these people >>>>> should be available also for final testing. Hostelers can also be a good >>>>> choice. >>>>> >>>>> # do we need to vary lighting condition (camera flash would give >>>>> constant light)? If yes, how ? >>>>> >>>>> # we have to photograph in white background. Maybe 6 chart papers will >>>>> form a perfect background. >>>>> >>>>> Can we do all that tomorrow? In my opinion, it can be done if we >>>>> delegate tasks as follows: >>>>> >>>>> 1. Abhishek TODO: bring camera >>>>> 2. Anjan TODO: arrange a white background (chart papers will do) >>>>> 3. Lakesh: coordinate everything as per your requirement >>>>> 4. Bibek: Contact your contacts and make arrangement. Bring camera >>>>> stand >>>>> >>>>> are you all ready? >>>>> >>>>> bye >>>>> >>>>> On Tue, Dec 2, 2008 at 7:07 PM, Anjan <roc...@gm...> wrote: >>>>> >>>>>> Lakesh's idea is good. But this is what i think. >>>>>> >>>>>> * The integration of the first phase is not yet complete. If the >>>>>> integration is done (i.e in the video sequence), the training of the >>>>>> recognition portion can be completed photographing in real world (eg. our >>>>>> classmates) will be easier, because the overhead of manual cropping will be >>>>>> removed, or at least reduced. So, the phase 1 integration in the video >>>>>> sequence is very necessary. We have to know Bibek's progress on that. >>>>>> >>>>>> * Instead of photographing our classmates, i think we should >>>>>> photograph other people. This is because, our friends will be unavailable >>>>>> later on when real testing is to be done. According to abhishek's idea of >>>>>> performing the testing explained in wiki (eg. in library hall) after our >>>>>> exam, our friends will be busy in their own projects and for the testing new >>>>>> photographs are to be recollected. So, i think instead of doing the job >>>>>> twice, we can photograph people who will be readily availabe (eg. third year >>>>>> guys etc). >>>>>> >>>>>> * At this point i think our plans are not much rigid. So, let's have a >>>>>> meeting soon (maybe tomorrow after DSP lab), and plan our schedules again. >>>>>> >>>>>> These are my opinions guys, let me know your opinion. >>>>>> >>>>>> On Tue, Dec 2, 2008 at 6:04 PM, Abhishek Dutta <adu...@gm...>wrote: >>>>>> >>>>>>> Hi all, >>>>>>> Lakesh was talking today about photographing our class mates and >>>>>>> performing some test (Lakesh can you explain in detail what you want to >>>>>>> do?). Lakesh has some free time so I think he can proceed with that testing >>>>>>> procedure. What do you guys think? >>>>>>> >>>>>>> Bye >>>>>>> >>>>>>> -- >>>>>>> Abhishek >>>>>>> http://adutta.np.googlepages.com >>>>>>> >>>>>>> >>>>>>> ------------------------------------------------------------------------- >>>>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>>>> challenge >>>>>>> Build the coolest Linux based applications with Moblin SDK & win >>>>>>> great prizes >>>>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>>>> world >>>>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>>>> _______________________________________________ >>>>>>> Rtftr-devel mailing list >>>>>>> Rtf...@li... >>>>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> Anjan Nepal >>>>>> >>>>>> GPO Box 11458 >>>>>> Registered Linux User #395963 >>>>>> >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------- >>>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>>> challenge >>>>>> Build the coolest Linux based applications with Moblin SDK & win great >>>>>> prizes >>>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>>> world >>>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>>> _______________________________________________ >>>>>> Rtftr-devel mailing list >>>>>> Rtf...@li... >>>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Abhishek >>>>> http://adutta.np.googlepages.com >>>>> >>>>> >>>>> ------------------------------------------------------------------------- >>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>> challenge >>>>> Build the coolest Linux based applications with Moblin SDK & win great >>>>> prizes >>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>> world >>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>> _______________________________________________ >>>>> Rtftr-devel mailing list >>>>> Rtf...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>> >>>>> >>>> >>>> >>>> -- >>>> Have a nice day >>>> >>> >>> >>> >>> -- >>> Abhishek >>> http://adutta.np.googlepages.com >>> >>> ------------------------------------------------------------------------- >>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>> challenge >>> Build the coolest Linux based applications with Moblin SDK & win great >>> prizes >>> Grand prize is a trip for two to an Open Source event anywhere in the >>> world >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>> _______________________________________________ >>> Rtftr-devel mailing list >>> Rtf...@li... >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>> >>> >> >> >> -- >> Anjan Nepal >> >> GPO Box 11458 >> Registered Linux User #395963 >> >> >> ------------------------------------------------------------------------- >> This SF.Net email is sponsored by the Moblin Your Move Developer's >> challenge >> Build the coolest Linux based applications with Moblin SDK & win great >> prizes >> Grand prize is a trip for two to an Open Source event anywhere in the >> world >> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >> _______________________________________________ >> Rtftr-devel mailing list >> Rtf...@li... >> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >> >> > > > -- > Have a nice day > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Abhishek http://adutta.np.googlepages.com |
|
From: lakesh <lak...@gm...> - 2008-12-03 01:22:28
|
+1 On Tue, Dec 2, 2008 at 9:32 PM, Anjan <roc...@gm...> wrote: > ACK > > > On Tue, Dec 2, 2008 at 9:07 PM, Abhishek Dutta <adu...@gm...>wrote: > >> Ok guys >> I am ready for clicking photos tomorrow. Moreover, I will also bring my >> camcorder so that we can have a video of people being photographed. We can >> use this video to do testing in later part. >> >> Everybody ACK if you are ready. >> >> bye >> >> >> On Tue, Dec 2, 2008 at 8:42 PM, lakesh <lak...@gm...> wrote: >> >>> Hi all, >>> What I actually want to test are the following. >>> 1) Most of the people, when they build a face detection and recognition >>> system use some threshold value to separate the in database and not in >>> database classification. This threshold directly depends on the training >>> sets. So for different sets different thresholds. So obviously there is >>> double work if we go for different data sets later. >>> 2) We haven't done geometric normalization which most of the people do. >>> Based on the paper Anjan referred we are going to take some detected images >>> of a person to train SubspaceLda. We don't know how this will perform. So I >>> want to check this as well >>> 3) I just want to make sure that we know where we are heading and be >>> prepared. We are using different research papers to implement the >>> algorithms. Specifically in the case of fr most of the papers they aren't >>> considering face detection part. They have just tested in the standard >>> databases. But in our case we are building an actual system where we have to >>> consider lots of things. JT sir said that it is ok even if we have 50% >>> accuracy. So it's a word of relief for us. However before going for video if >>> we can just test using the still images we will get some hints I guess. >>> >>> >>> On Tue, Dec 2, 2008 at 8:25 PM, Abhishek Dutta <adu...@gm...>wrote: >>> >>>> Hi all, >>>> I agree to all Anjan's points. Phase 1 integration is necessary. And >>>> lakesh seems to be waiting for bibek. >>>> >>>> Moreover Lakesh want to photograph subjects right now to find the >>>> "threshold" values (I guess). So 10 subjects with 10 pose variation means >>>> total 100 photos. This is not a huge task. I can bring my camera and bibek >>>> will bring his stand. We can then photograph 10 people. Now we have to >>>> decide the following stuffs >>>> >>>> # which 10 people: I am also not in favor of photographing our class >>>> mates as they might be busy during final testing phase (Jan 01, 2009 to Jan >>>> 15, 2009). So we have to find volunteers. Bibek has lots of contacts from >>>> 3rd year. He can arrange that. We have to keep in mind that these people >>>> should be available also for final testing. Hostelers can also be a good >>>> choice. >>>> >>>> # do we need to vary lighting condition (camera flash would give >>>> constant light)? If yes, how ? >>>> >>>> # we have to photograph in white background. Maybe 6 chart papers will >>>> form a perfect background. >>>> >>>> Can we do all that tomorrow? In my opinion, it can be done if we >>>> delegate tasks as follows: >>>> >>>> 1. Abhishek TODO: bring camera >>>> 2. Anjan TODO: arrange a white background (chart papers will do) >>>> 3. Lakesh: coordinate everything as per your requirement >>>> 4. Bibek: Contact your contacts and make arrangement. Bring camera stand >>>> >>>> are you all ready? >>>> >>>> bye >>>> >>>> On Tue, Dec 2, 2008 at 7:07 PM, Anjan <roc...@gm...> wrote: >>>> >>>>> Lakesh's idea is good. But this is what i think. >>>>> >>>>> * The integration of the first phase is not yet complete. If the >>>>> integration is done (i.e in the video sequence), the training of the >>>>> recognition portion can be completed photographing in real world (eg. our >>>>> classmates) will be easier, because the overhead of manual cropping will be >>>>> removed, or at least reduced. So, the phase 1 integration in the video >>>>> sequence is very necessary. We have to know Bibek's progress on that. >>>>> >>>>> * Instead of photographing our classmates, i think we should photograph >>>>> other people. This is because, our friends will be unavailable later on when >>>>> real testing is to be done. According to abhishek's idea of performing the >>>>> testing explained in wiki (eg. in library hall) after our exam, our friends >>>>> will be busy in their own projects and for the testing new photographs are >>>>> to be recollected. So, i think instead of doing the job twice, we can >>>>> photograph people who will be readily availabe (eg. third year guys etc). >>>>> >>>>> * At this point i think our plans are not much rigid. So, let's have a >>>>> meeting soon (maybe tomorrow after DSP lab), and plan our schedules again. >>>>> >>>>> These are my opinions guys, let me know your opinion. >>>>> >>>>> On Tue, Dec 2, 2008 at 6:04 PM, Abhishek Dutta <adu...@gm...>wrote: >>>>> >>>>>> Hi all, >>>>>> Lakesh was talking today about photographing our class mates and >>>>>> performing some test (Lakesh can you explain in detail what you want to >>>>>> do?). Lakesh has some free time so I think he can proceed with that testing >>>>>> procedure. What do you guys think? >>>>>> >>>>>> Bye >>>>>> >>>>>> -- >>>>>> Abhishek >>>>>> http://adutta.np.googlepages.com >>>>>> >>>>>> >>>>>> ------------------------------------------------------------------------- >>>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>>> challenge >>>>>> Build the coolest Linux based applications with Moblin SDK & win great >>>>>> prizes >>>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>>> world >>>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>>> _______________________________________________ >>>>>> Rtftr-devel mailing list >>>>>> Rtf...@li... >>>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Anjan Nepal >>>>> >>>>> GPO Box 11458 >>>>> Registered Linux User #395963 >>>>> >>>>> >>>>> >>>>> ------------------------------------------------------------------------- >>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>> challenge >>>>> Build the coolest Linux based applications with Moblin SDK & win great >>>>> prizes >>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>> world >>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>> _______________________________________________ >>>>> Rtftr-devel mailing list >>>>> Rtf...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>> >>>>> >>>> >>>> >>>> -- >>>> Abhishek >>>> http://adutta.np.googlepages.com >>>> >>>> >>>> ------------------------------------------------------------------------- >>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>> challenge >>>> Build the coolest Linux based applications with Moblin SDK & win great >>>> prizes >>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>> world >>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>> _______________________________________________ >>>> Rtftr-devel mailing list >>>> Rtf...@li... >>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>> >>>> >>> >>> >>> -- >>> Have a nice day >>> >> >> >> >> -- >> Abhishek >> http://adutta.np.googlepages.com >> >> ------------------------------------------------------------------------- >> This SF.Net email is sponsored by the Moblin Your Move Developer's >> challenge >> Build the coolest Linux based applications with Moblin SDK & win great >> prizes >> Grand prize is a trip for two to an Open Source event anywhere in the >> world >> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >> _______________________________________________ >> Rtftr-devel mailing list >> Rtf...@li... >> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >> >> > > > -- > Anjan Nepal > > GPO Box 11458 > Registered Linux User #395963 > > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Have a nice day |
|
From: Anjan <roc...@gm...> - 2008-12-02 15:47:56
|
ACK On Tue, Dec 2, 2008 at 9:07 PM, Abhishek Dutta <adu...@gm...> wrote: > Ok guys > I am ready for clicking photos tomorrow. Moreover, I will also bring my > camcorder so that we can have a video of people being photographed. We can > use this video to do testing in later part. > > Everybody ACK if you are ready. > > bye > > > On Tue, Dec 2, 2008 at 8:42 PM, lakesh <lak...@gm...> wrote: > >> Hi all, >> What I actually want to test are the following. >> 1) Most of the people, when they build a face detection and recognition >> system use some threshold value to separate the in database and not in >> database classification. This threshold directly depends on the training >> sets. So for different sets different thresholds. So obviously there is >> double work if we go for different data sets later. >> 2) We haven't done geometric normalization which most of the people do. >> Based on the paper Anjan referred we are going to take some detected images >> of a person to train SubspaceLda. We don't know how this will perform. So I >> want to check this as well >> 3) I just want to make sure that we know where we are heading and be >> prepared. We are using different research papers to implement the >> algorithms. Specifically in the case of fr most of the papers they aren't >> considering face detection part. They have just tested in the standard >> databases. But in our case we are building an actual system where we have to >> consider lots of things. JT sir said that it is ok even if we have 50% >> accuracy. So it's a word of relief for us. However before going for video if >> we can just test using the still images we will get some hints I guess. >> >> >> On Tue, Dec 2, 2008 at 8:25 PM, Abhishek Dutta <adu...@gm...>wrote: >> >>> Hi all, >>> I agree to all Anjan's points. Phase 1 integration is necessary. And >>> lakesh seems to be waiting for bibek. >>> >>> Moreover Lakesh want to photograph subjects right now to find the >>> "threshold" values (I guess). So 10 subjects with 10 pose variation means >>> total 100 photos. This is not a huge task. I can bring my camera and bibek >>> will bring his stand. We can then photograph 10 people. Now we have to >>> decide the following stuffs >>> >>> # which 10 people: I am also not in favor of photographing our class >>> mates as they might be busy during final testing phase (Jan 01, 2009 to Jan >>> 15, 2009). So we have to find volunteers. Bibek has lots of contacts from >>> 3rd year. He can arrange that. We have to keep in mind that these people >>> should be available also for final testing. Hostelers can also be a good >>> choice. >>> >>> # do we need to vary lighting condition (camera flash would give constant >>> light)? If yes, how ? >>> >>> # we have to photograph in white background. Maybe 6 chart papers will >>> form a perfect background. >>> >>> Can we do all that tomorrow? In my opinion, it can be done if we delegate >>> tasks as follows: >>> >>> 1. Abhishek TODO: bring camera >>> 2. Anjan TODO: arrange a white background (chart papers will do) >>> 3. Lakesh: coordinate everything as per your requirement >>> 4. Bibek: Contact your contacts and make arrangement. Bring camera stand >>> >>> are you all ready? >>> >>> bye >>> >>> On Tue, Dec 2, 2008 at 7:07 PM, Anjan <roc...@gm...> wrote: >>> >>>> Lakesh's idea is good. But this is what i think. >>>> >>>> * The integration of the first phase is not yet complete. If the >>>> integration is done (i.e in the video sequence), the training of the >>>> recognition portion can be completed photographing in real world (eg. our >>>> classmates) will be easier, because the overhead of manual cropping will be >>>> removed, or at least reduced. So, the phase 1 integration in the video >>>> sequence is very necessary. We have to know Bibek's progress on that. >>>> >>>> * Instead of photographing our classmates, i think we should photograph >>>> other people. This is because, our friends will be unavailable later on when >>>> real testing is to be done. According to abhishek's idea of performing the >>>> testing explained in wiki (eg. in library hall) after our exam, our friends >>>> will be busy in their own projects and for the testing new photographs are >>>> to be recollected. So, i think instead of doing the job twice, we can >>>> photograph people who will be readily availabe (eg. third year guys etc). >>>> >>>> * At this point i think our plans are not much rigid. So, let's have a >>>> meeting soon (maybe tomorrow after DSP lab), and plan our schedules again. >>>> >>>> These are my opinions guys, let me know your opinion. >>>> >>>> On Tue, Dec 2, 2008 at 6:04 PM, Abhishek Dutta <adu...@gm...>wrote: >>>> >>>>> Hi all, >>>>> Lakesh was talking today about photographing our class mates and >>>>> performing some test (Lakesh can you explain in detail what you want to >>>>> do?). Lakesh has some free time so I think he can proceed with that testing >>>>> procedure. What do you guys think? >>>>> >>>>> Bye >>>>> >>>>> -- >>>>> Abhishek >>>>> http://adutta.np.googlepages.com >>>>> >>>>> >>>>> ------------------------------------------------------------------------- >>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>>> challenge >>>>> Build the coolest Linux based applications with Moblin SDK & win great >>>>> prizes >>>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>>> world >>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>>> _______________________________________________ >>>>> Rtftr-devel mailing list >>>>> Rtf...@li... >>>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>>> >>>>> >>>> >>>> >>>> -- >>>> Anjan Nepal >>>> >>>> GPO Box 11458 >>>> Registered Linux User #395963 >>>> >>>> >>>> >>>> ------------------------------------------------------------------------- >>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>> challenge >>>> Build the coolest Linux based applications with Moblin SDK & win great >>>> prizes >>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>> world >>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>> _______________________________________________ >>>> Rtftr-devel mailing list >>>> Rtf...@li... >>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>> >>>> >>> >>> >>> -- >>> Abhishek >>> http://adutta.np.googlepages.com >>> >>> ------------------------------------------------------------------------- >>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>> challenge >>> Build the coolest Linux based applications with Moblin SDK & win great >>> prizes >>> Grand prize is a trip for two to an Open Source event anywhere in the >>> world >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>> _______________________________________________ >>> Rtftr-devel mailing list >>> Rtf...@li... >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>> >>> >> >> >> -- >> Have a nice day >> > > > > -- > Abhishek > http://adutta.np.googlepages.com > > ------------------------------------------------------------------------- > This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge > Build the coolest Linux based applications with Moblin SDK & win great > prizes > Grand prize is a trip for two to an Open Source event anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > Rtftr-devel mailing list > Rtf...@li... > https://lists.sourceforge.net/lists/listinfo/rtftr-devel > > -- Anjan Nepal GPO Box 11458 Registered Linux User #395963 |
|
From: Abhishek D. <adu...@gm...> - 2008-12-02 15:22:43
|
Ok guys I am ready for clicking photos tomorrow. Moreover, I will also bring my camcorder so that we can have a video of people being photographed. We can use this video to do testing in later part. Everybody ACK if you are ready. bye On Tue, Dec 2, 2008 at 8:42 PM, lakesh <lak...@gm...> wrote: > Hi all, > What I actually want to test are the following. > 1) Most of the people, when they build a face detection and recognition > system use some threshold value to separate the in database and not in > database classification. This threshold directly depends on the training > sets. So for different sets different thresholds. So obviously there is > double work if we go for different data sets later. > 2) We haven't done geometric normalization which most of the people do. > Based on the paper Anjan referred we are going to take some detected images > of a person to train SubspaceLda. We don't know how this will perform. So I > want to check this as well > 3) I just want to make sure that we know where we are heading and be > prepared. We are using different research papers to implement the > algorithms. Specifically in the case of fr most of the papers they aren't > considering face detection part. They have just tested in the standard > databases. But in our case we are building an actual system where we have to > consider lots of things. JT sir said that it is ok even if we have 50% > accuracy. So it's a word of relief for us. However before going for video if > we can just test using the still images we will get some hints I guess. > > > On Tue, Dec 2, 2008 at 8:25 PM, Abhishek Dutta <adu...@gm...>wrote: > >> Hi all, >> I agree to all Anjan's points. Phase 1 integration is necessary. And >> lakesh seems to be waiting for bibek. >> >> Moreover Lakesh want to photograph subjects right now to find the >> "threshold" values (I guess). So 10 subjects with 10 pose variation means >> total 100 photos. This is not a huge task. I can bring my camera and bibek >> will bring his stand. We can then photograph 10 people. Now we have to >> decide the following stuffs >> >> # which 10 people: I am also not in favor of photographing our class mates >> as they might be busy during final testing phase (Jan 01, 2009 to Jan 15, >> 2009). So we have to find volunteers. Bibek has lots of contacts from 3rd >> year. He can arrange that. We have to keep in mind that these people should >> be available also for final testing. Hostelers can also be a good choice. >> >> # do we need to vary lighting condition (camera flash would give constant >> light)? If yes, how ? >> >> # we have to photograph in white background. Maybe 6 chart papers will >> form a perfect background. >> >> Can we do all that tomorrow? In my opinion, it can be done if we delegate >> tasks as follows: >> >> 1. Abhishek TODO: bring camera >> 2. Anjan TODO: arrange a white background (chart papers will do) >> 3. Lakesh: coordinate everything as per your requirement >> 4. Bibek: Contact your contacts and make arrangement. Bring camera stand >> >> are you all ready? >> >> bye >> >> On Tue, Dec 2, 2008 at 7:07 PM, Anjan <roc...@gm...> wrote: >> >>> Lakesh's idea is good. But this is what i think. >>> >>> * The integration of the first phase is not yet complete. If the >>> integration is done (i.e in the video sequence), the training of the >>> recognition portion can be completed photographing in real world (eg. our >>> classmates) will be easier, because the overhead of manual cropping will be >>> removed, or at least reduced. So, the phase 1 integration in the video >>> sequence is very necessary. We have to know Bibek's progress on that. >>> >>> * Instead of photographing our classmates, i think we should photograph >>> other people. This is because, our friends will be unavailable later on when >>> real testing is to be done. According to abhishek's idea of performing the >>> testing explained in wiki (eg. in library hall) after our exam, our friends >>> will be busy in their own projects and for the testing new photographs are >>> to be recollected. So, i think instead of doing the job twice, we can >>> photograph people who will be readily availabe (eg. third year guys etc). >>> >>> * At this point i think our plans are not much rigid. So, let's have a >>> meeting soon (maybe tomorrow after DSP lab), and plan our schedules again. >>> >>> These are my opinions guys, let me know your opinion. >>> >>> On Tue, Dec 2, 2008 at 6:04 PM, Abhishek Dutta <adu...@gm...>wrote: >>> >>>> Hi all, >>>> Lakesh was talking today about photographing our class mates and >>>> performing some test (Lakesh can you explain in detail what you want to >>>> do?). Lakesh has some free time so I think he can proceed with that testing >>>> procedure. What do you guys think? >>>> >>>> Bye >>>> >>>> -- >>>> Abhishek >>>> http://adutta.np.googlepages.com >>>> >>>> >>>> ------------------------------------------------------------------------- >>>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>>> challenge >>>> Build the coolest Linux based applications with Moblin SDK & win great >>>> prizes >>>> Grand prize is a trip for two to an Open Source event anywhere in the >>>> world >>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>> _______________________________________________ >>>> Rtftr-devel mailing list >>>> Rtf...@li... >>>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>>> >>>> >>> >>> >>> -- >>> Anjan Nepal >>> >>> GPO Box 11458 >>> Registered Linux User #395963 >>> >>> >>> ------------------------------------------------------------------------- >>> This SF.Net email is sponsored by the Moblin Your Move Developer's >>> challenge >>> Build the coolest Linux based applications with Moblin SDK & win great >>> prizes >>> Grand prize is a trip for two to an Open Source event anywhere in the >>> world >>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>> _______________________________________________ >>> Rtftr-devel mailing list >>> Rtf...@li... >>> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >>> >>> >> >> >> -- >> Abhishek >> http://adutta.np.googlepages.com >> >> ------------------------------------------------------------------------- >> This SF.Net email is sponsored by the Moblin Your Move Developer's >> challenge >> Build the coolest Linux based applications with Moblin SDK & win great >> prizes >> Grand prize is a trip for two to an Open Source event anywhere in the >> world >> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >> _______________________________________________ >> Rtftr-devel mailing list >> Rtf...@li... >> https://lists.sourceforge.net/lists/listinfo/rtftr-devel >> >> > > > -- > Have a nice day > -- Abhishek http://adutta.np.googlepages.com |
23 messages has been excluded from this view by a project administrator.
