Another question, I was looking at the wordnet extension, does this only work with Matlab also

Paul

Um...I dont use matlab :-), or should I say I do not have matlab

Paul

Currently, the annotation tool does not allow suggesting a default set of labels. The user must type in the label.

To interact with the data, there is a Matlab toolbox with various functions for plotting, extracting a polygon mask, etc. You may find these functions helpful for parsing and extracting various properties (e.g. points interior to the annotated polygon).

The Matlab toolbox can be downloaded here:

http://labelme.csail.mit.edu/LabelMeToolbox/index.html

Guys tks for the help on the install, it seems to have been a file owner problem.

So now I have a few more questions :-)

1. When I annotate a image, and it asks me to label it, is it possible to suggest a default set of labels from somewhere

2. When the annotation is done it stores them in the Annotations folder, so I had a look at the file, but wanted to know what the fields were: I have

<annotation><filename>boo.jpg</filename><folder>folder</folder><source><sourceImage>The MIT-CSAIL database of objects and scenes</sourceImage><sourceAnnotation>LabelMe Webtool</sourceAnn
otation></source><object><name>laptop</name><deleted>0</deleted><verified>0</verified><date>13-Nov-2009 16:11:14</date><id>0</id><polygon><username>anonymous</username><pt><x>37</x><y>348</y></pt><pt>
<x>47</x><y>451</y></pt><pt><x>108</x><y>611</y></pt><pt><x>190</x><y>776</y></pt><pt><x>382</x><y>858</y></pt><pt><x>1153</x><y>677</y></pt><pt><x>1072</x><y>424</y></pt><pt><x>933</x><y>279</y></
pt><pt><x>532</x><y>296</y></pt></polygon></object><imagesize><nrows>960
</nrows><ncols>1280
</ncols></imagesize></annotation>

Is the x/y where I clicked as I went around the picture, or some other points. Also if I know the x,y points of the picture as given by the data above, does anyone know of a simple way of storing the pixels contained within

e.g I have a square shape x1=10, y1=10, x2=20, y2=10, x3=20, y3=20, x4=10, y4=20. Now this is what it stored in the annotation file from what I can see. However if I knew a point say 15,15, how can I work out that that is within the square, I know there must be a formula for this somewhere, just cant seet it :-)

tks

Paul

Done, and tks

Please e-mail me directly (russell AT di DOT ens DOT fr) the URL. I'll have a look. Thanks.

Is there a way of posting the url offline ?

Also apache contained the perl/CGI things you mentioned. Also in the apache logfiles I am seeing

Premature end of script headers: write_logfile.cgi, referer: http://www.mydomain.com/annotate/LabelMe-1...der=null&image=null

and alot of the other scripts like get_anno_file.cgi also showing the same "null" error

I also get

File does not exist: /usr/local/mydomain/annotate/LabelMe-1-99/Images/null, referer: http://www.mydomain.com/annotate/LabelMe-1...der=null&image=null

BUt still getting the same objXml.status error, and no image loading

Hi Paul,

Can you please send me the URL of your copy of the annotation tool? I can have a quick look.

Make sure that you are enabling perl/CGI in your httpd.conf file:

Options +Includes +ExecCGI
AddType text/html .shtml
AddOutputFilter INCLUDES .shtml
AddHandler cgi-script .cgi .pl

Hi

I downloaded the latest version and went through the install instructions (ubuntu), created the 3 directories, and ran make, which all went well. Then I went to the URL

www.myname.com/tool.html, and the error unknown objXml.status pops up a couple of times, and then the icons appear, BUT no image loaded.

I am seeing alot of "premature end of script headers" in my apache error_log also.

I am running Firefox on a mac, and have checked the Page info for xhtml +xml, and it shows the correct info.

Any pointers...

Paul

thanks for pointing out this bug. Please, download the toolbox again from the labelme website. That bug should be fixed now. Let us know if you still encounter any problem.


On Nov 3, 2009, at 6:24 AM, QT - Gibbs wrote:

< replied-to message removed by QT >

Hi guys,
I think you may have a tiny bug.
I've just download the last LabelMe toolbox, when I tried to run the following lines I got an error

>> HOMEANNOTATIONS = 'http://labelme.csail.mit.edu/Annotations';
>> HOMEIMAGES = 'http://labelme.csail.mit.edu/Images';
>> D = LMdatabase(HOMEANNOTATIONS, {'static_street_statacenter_cambridge_outdoor_2005'});
??? Undefined function or variable "Nfolders".

Error in ==> LMdatabase at 109
for f = 1:Nfolders

------------
It was easily solved when I added in line 70 of LMdatabase.m:
    if Narg == 2
        Folder = varargin{2}
        Nfolders = length(Folder)
    end

I don't know if it's the right solution but it worked for me.

Hi, I used the scene search function from the web interface to retrieve the "cross street", and get 325 images returned. However, when I use the matlab code to do the retrieval, I get nothing. The code I used is like this:
HOMEIMAGES = 'http://people.csail.mit.edu/brussell/research/LabelMe/Images';
HOMEANNOTATIONS = 'http://people.csail.mit.edu/brussell/research/LabelMe/Annotations';
D = LMdatabase(HOMEANNOTATIONS);
[Q,j] = LMquery(D, 'scenedescription', 'cross street');

My question is: how can I get the same search result as the web interface?

I have downloaded both Label Me and Annotation Tool box for scene perception experiment. Based on the instruction, all data files are saved in Annotation folder in XML format. If we are running the same experiment for many subjects and we would like to see each participant's capability in labelling the object on the scene. How can I modify the matlab program (maybe Label Me or AnnotationTool) to achieve the following purposes:
1) Save the annotation data for each subject in separate folders for analysis purpose
2) Read the result from the XML as an image to scoring purpose (e.g. whether they label the right object at the right location)?
Many thanks for your help!