Box It is a web app for annotating images. This version includes license plate specific features.
This app takes advantage of unusual browser features, so support may spotty.
Currently tested on Chrome 69.0 and Firefox 62.0. Chrome works best.
Clone this project and launch with a web server (such as caddy), or simply open https://bigmlcom.github.io/boxit/.
The examples
and examples/images
directory contains example
annotation files and example images, respectively.
The Select Images
button allows a directory of images to be loaded
for annotation. While there may be dialog confirming the upload of the
images to the server, the images are not shared with the server. The
selected files are only accessed by the browser.
The right and left arrow keys will step through the selected images. Alternatively the top slider can be used to jump to an image.
The view window's zoom level may be controlled with the mouse wheel, or double-clicking to zoom. The view window also supports panning through a click and drag.
After selecting an annotation mode and class (see below), an region may be annotated by simply clicking once to start an annotation box and a second time to finalize the box. A box can be deleted with backspace. If there are multiple boxes, they will be removed in the order they were created.
To give the entire image a class label, either click the Set Image Class
button or press the enter
key.
This version of Box It has two annotation modes.
First is the custom
mode, in which a user creates their own set of
annotation classes. Each annotation class may be selected by either
clicking on the corresponding button or by using keys 0-9. The number
keys will select the annotations according the order they appear in
the sidebar.
Second is the license plate
mode. This mode comes with alpha-numeric
annotations predefined. Hotkeys are simply the corresponding keys on
the keyboard. Note that 0
and O
are both interpreted as 0
.
When loading an annotation file, the mode will be autodetected.
Annotations are cached with the browser, so closing the browser will
not cause your work to be lost. However, the Clear Annotations
button will delete the cached annotations.
The Save Annotations
button will save the annotations in a JSON
format. This will create an annotations.json
file in your browser's
default downloads directory.
Similarly, the Load Annotations
button will load annotations from a
JSON file. The loaded annotations will override any existing or cached
annotations in the browser.
The Extract Images
button will crop and save each boxed region as a
seperate file in your browser's default download directory. This is
especially useful for annotating heirarchial objects that will be
modeled seperately.
Copyright (C) 2019 BigML Inc.
Distributed under the Apache License, Version 2.0.