Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Source code for the LabelMe annotation tool.
C HTML JavaScript C++ Shell Objective-C Other
Branch: master

README.md

LabelMe annotation tool source code

Here you will find the source code to install the LabelMe annotation tool on your server. LabelMe is an annotation tool writen in Javascript for online image labeling. The advantage with respect to traditional image annotation tools is that you can access the tool from anywhere and people can help you to annotate your images without having to install or copy a large dataset onto their computers.

CITATION:

B. C. Russell, A. Torralba, K. P. Murphy, W. T. Freeman. LabelMe: a Database and Web-based Tool for Image Annotation. International Journal of Computer Vision, 77(1-3):157-173, 2008. Project page

DOWNLOAD:

You can download a zip file of the source code directly.

Alternatively, you can clone it from GitHub as follows:

$ git clone https://github.com/CSAILVision/LabelMeAnnotationTool.git

CONTENTS:

  • Images - This is where your images go.
  • Annotations - This is where the annotations are collected.
  • Masks - This is where the segmentation masks are collected (scribble mode).
  • Scribbles - This is where the scribbles are collected (scribble mode).
  • tool.html - Main web page for LabelMe annotation tool.
  • annotationTools - Directory with source code.
  • annotationCache - Location of temporary files.
  • Icons - Icons used on web page.

QUICK START INSTRUCTIONS:

  1. Put LabelMe annotation tool code on web server (see web server configuration requirements below).

  2. On the command line run:

    $ make

    Alternatively, if you want to install only the basic tool and none of the advanced features, such as scribbles, run on the command line instead:

    $ make basic

    This will set a global variable that the perl scripts need. Note If you move the location of the code, then you need to re-run "make" to refresh the global variable.

  3. Create a subfolder inside the "Images" folder and place your images there. For example: "Images/example_folder/img1.jpg". Make sure all of your images have a ".jpg" extension and the folders/filenames have alphanumeric characters (i.e. no spaces or funny characters).

  4. Point your web browser to the following URL:

    http://www.yourserver.edu/path/to/LabelMe/tool.html?collection=LabelMe&mode=f&folder=example_folder&image=img1.jpg

  5. Label your image. Press "show me another image" to go to the next image in the folder.

  6. Voila! Your annotations will appear inside of the "Annotations" folder.

WEB SERVER REQUIREMENTS:

You will need the following to set up the LabelMe tool on your web server:

  • Run an Apache server (see special configuration instructions for Ubuntu or Windows).
  • Enable authconfig in Apache so that server side includes (SSI) will work. This will allow SVG drawing capabilities. This is the most common source of errors, so make sure this step is working.
  • Allow perl/CGI scripts to run. This is the second most common source of errors.
  • Make sure the php5 and libapache2-mod-php5 libraries are installed. You can install them on Linux by running the following:

    $ sudo apt-get install php5 libapache2-mod-php5
  • (Optional) See special configuration instructions if you are installing on Ubuntu or Windows.

If you are not able to draw polygons, check to see if the page is loaded as an "application/xhtml+xml" page (you can see this in Firefox by navigating to Tools->Page Info). If it is not, be sure that SSI are enabled (see above for enabling authconfig in Apache).

Make sure that your images have read permissions on your web server and folders in the "Annotations" folder have write permissions. Also, "annotationCache/TmpAnnotations" needs to have write permissions.

FEATURES OF THE ANNOTATION TOOL:

  • The following are URL variables you can pass to the annotation tool:

    • mode=im - Only show the image and drawing canvas (do not show anything outside of the image.
    • mode=mt - Mechanical Turk mode.
    • mode=f - Pressing "next image" button goes to next image in the folder.
    • mode=i - Pressing "next image" button goes to random image in the default LabelMe collection.
    • mode=c - Go to next image in the collection (set via the dirlist).
    • username=johndoe - Sets username for labeling session.
    • collection=LabelMe - Uses the default LabelMe collection list. See below for setting up a new collection list.
    • folder=MyLabelMeFolder - LabelMe folder where the image lives.
    • image=image.jpg - LabelMe image to annotate.
    • objects=car,person,building - When popup bubble appears asking the user for the object name, the user selects one of these objects appearing as a drop-down list.
    • scribble=false - Turns off scribble mode.
    • objlist=visible - This controls whether the object list on the right side is visible or not. Use "objlist=hidden" to make it hidden.
    • actions=n - Control what actions the user is allowed to do. To set the desired actions, use any combination of the letters below. For example, to allow renaming, modify control points, and delete actions, then set "actions=rmd". By default, "actions=n". The following are possible actions:
      • n - create and edit new polygons
      • r - rename existing objects
      • m - modify control points on existing objects
      • d - delete existing objects
      • a - allow all actions
      • v - view polygons only (do not allow any editing)
    • viewobj=e - Control which objects the user sees. Use one of the following possible options below. By default, "viewobj=e". Note that for deleted objects, these will be shown in gray and the object name in the object list will be italicized.
      • e - view new and previously labeled objects
      • n - view new objects only
      • d - view new and deleted objects
      • a - view all objects (new, existing, deleted)

    The following are for Mechanical Turk mode:

    • mt_sandbox=true - Use Mechanical Turk sandbox mode. This mode is used for debugging on Mechanical Turk. You may want to start with this variable set to make sure everything works.
    • N=5 - The worker is required to label at least 5 polygons. Use N=inf to allow the worker to label as many as they want.
    • mt_intro=http://yourpage.com - You may customize the instructions that the worker sees. By default, the following instructions are given to the workers.
    • mt_instructions=Place your instructions here - You may customize the one-line instructions that the worker sees at the top of the labeling task. By default, the instructions are: Please label as many objects as you want in this image.
  • You can create a collection of images to label by running the following on the command line:

    $ cd ./annotationTools/sh/
    $ ./populate_dirlist.sh

    This will create a list of all images inside the "./Images" folder, and will appear inside the file "./annotationCache/DirLists/labelme.txt".

    You can then label images inside the collection using the following URL:

    http://www.yourserver.edu/path/to/LabelMe/tool.html?collection=labelme&mode=i

    You can create a collection consisting of a particular folder by running the following from the command line:

    $ cd ./annotationTools/sh/
    $ ./populate_dirlist.sh my_collection.txt example_folder

    The list will appear inside "./annotationCache/DirLists/my_collection.txt". You can then label images inside the collection using the following URL:

    http://www.yourserver.edu/path/to/LabelMe/tool.html?collection=my_collection&mode=i

  • You can change the layout of the annotation files for your collection by modifying the XML file template inside of "./annotationCache/XMLTemplates/your_collection.xml". The default template is "./annotationCache/XMLTemplates/labelme.xml".

  • A log file of the annotation tool actions are recorded in "./annotationCache/Logs/logfile.txt". Make sure that this file has write permissions.

CODE API

The following is a brief overview of the source code. Please see the Javascript code API for more details.

  • tool.html - This is the entry point for the annotation tool. The main functionality is to insert all of the javascript code and lay down the drawing canvases.

  • annotationTools/js/ - This folder contains all of the javascript code for the annotation tool functionalities. We provide the code API for the Javascript source code, which has been automatically extracted from the source code comments.

  • annotationTools/perl/ - This folder contains all of the Perl scripts used for communication with the server back-end.

  • annotationTools/css/ - This folder contains all of the CSS style definitions.

  • annotationTools/html/ - This folder contains auxillary HTML files (e.g. for Mechanical Turk instructions, etc.).


(c) 2015, MIT Computer Science and Artificial Intelligence Laboratory

Something went wrong with that request. Please try again.