README.md

Bounding box annotator

A bounding box annotation tool written in CoffeeScript/JavaScript. There is an online demo at https://kyamagu.github.io/bbox-annotator/demo.html

Contents

bbox_annotator.coffee  Main code.
demo.html              Example to host on a private server.
mturk.html             Example to use in Amazon MTurk.
README.md              This documentation.

Be sure to unescape × with × when directly editing an MTurk template on the Amazon requester UI.

Compile

Compile the coffee script to get a javascript.

coffee -c bbox_annotator.coffee

Or, go to http://coffeescript.org/ and compile the coffeescript.

Usage

Load the compiled javascript codes inside HTML.

<script src="bbox_annotator.js" type="text/javascript"></script>

Embed a block element for annotation tool and a form element in HTML.

<div id="bbox_annotator"></div>
<input id="annotation_data" name="annotation_data" type="hidden" />

Then, attach a piece of script to launch the annotation tool. The annotator takes a callback function on change of the annotation data. Use JSON format to display the value.

<script type="text/javascript">
$(document).ready(function() {
  var editor = new BBoxAnnotator({
    url: "/url/to/image.jpg",
    onchange: function(annotation) {
      $("#annotation_data").val(JSON.stringify(annotation));
    }
  });
});
</script>

The returned annotation is an array of objects. Here is an example.

[
  {
    "top": 0,
    "left": 0,
    "width": 100,
    "height": 100,
    "label", "object"
  }
]

Check the attached demo.html and mturk.html for how to use.

Options

The BBoxAnnotator object takes options to change its behavior.

• id: CSS selector for the annotator element. Default is "#bbox_annotator".
• input_method: One of the "text", "select", and "fixed". This will change how to input associated annotation text.
• labels: A string, or an array of strings. Set of labels used for suggestion in "text" or "select" input. When fixed, only the first element is chosen.
• width: Width of the image. The default is the original width of the image.
• height: Height of the image. The default is the original height of the image.
• show_label: Flag to display a label in the box. Default is true for "text" and "select" input, and false for "fixed" input.
• border_width: Size of border around the image. Default is 2. Leaving it 0 can make it hard to start annotation around the image border.
• multiple: Defines if multiple boxes could be selected. Default is true.
• guide: Enable vertical and horizontal guide lines. Specify true, or object that specifies color like {'color': '#fff'}.

License

The code may be redistributed under BSD license.