<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Lukas Sanner on Medium]]></title>
        <description><![CDATA[Stories by Lukas Sanner on Medium]]></description>
        <link>https://medium.com/@lukas.sanner?source=rss-d7fed06405c7------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*byC2-amWfyxn4ZuxaGBk4Q.jpeg</url>
            <title>Stories by Lukas Sanner on Medium</title>
            <link>https://medium.com/@lukas.sanner?source=rss-d7fed06405c7------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:26:10 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lukas.sanner/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Magic Wand Annotation Tool]]></title>
            <link>https://medium.com/moonvision/magic-wand-annotation-tool-a73f26acde21?source=rss-d7fed06405c7------2</link>
            <guid isPermaLink="false">https://medium.com/p/a73f26acde21</guid>
            <category><![CDATA[vienna]]></category>
            <category><![CDATA[annotation-tools]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[machine-vision]]></category>
            <category><![CDATA[spectral-clustering]]></category>
            <dc:creator><![CDATA[Lukas Sanner]]></dc:creator>
            <pubDate>Tue, 09 Apr 2019 08:25:14 GMT</pubDate>
            <atom:updated>2019-04-10T07:23:55.972Z</atom:updated>
            <content:encoded><![CDATA[<h3>Computer Vision: Magic Wand Annotation Tool</h3><p>For supervised machine learning it is necessary to have precisely labelled data. Annotating data is a tedious and cumbersome task, especially when it comes to a massive amount of items to label with a complex structure respectively border.</p><p>For example, I want a segmentation network for dogs, there are different kinds of dogs and for every frame they will have a new pose.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/992/0*a_c6fT3fZto2LDG0" /></figure><p>Figure 1: Source: <a href="http://host.robots.ox.ac.uk/pascal/VOC/">the PASCAL VOC project</a></p><p>To make things easier and gain speed while annotating we created a magic wand annotation tool which helps you finding complex borders of an object just by setting a rectangle-like region of interest (ROI).</p><h3>Implementation</h3><p>For the magic wand we compare the pixels in the image in terms of similarity, thus we can group similar pixels by <a href="https://arxiv.org/pdf/0711.0189.pdf">spectral clustering</a> and create a contour of the object.</p><p>There are some questions that arise when <strong>comparing pixels</strong>, one can compare pixels by:</p><ul><li>RGB features</li><li>Light intensity</li><li>Texture</li><li>Positional coordinates</li><li>Hue</li></ul><p>In addition, if you compare every pixel with every other pixel a massive operator is created, e.g. for a ROI of 300x300 pixels, one creates a <strong>90,000x90,000 similarity matrix, </strong>a symmetric matrix which describes the similarity of every pixel to every other<strong>,</strong> which requires a considerable amount of computation resources and memory.</p><p>So we have to decide which features do we compare and how to deal with that amount of data.</p><p>After some research we ended up by using <strong>a combination of RGB values and coordinates </strong>and implemented an <a href="http://www.math.kent.edu/~reichel/publications/aca.pdf">approximation scheme</a> of the large similarity matrix, which reduces the amount of data to be processed, so we can compute the cluster in real time. In addition we use a <a href="http://papers.nips.cc/paper/2619-self-tuning-spectral-clustering.pdf">self learning similarity metric</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*odpBbkxqgVJD168F" /></figure><p>Figure 2: from left to right, original image, drawn ROI, computed contour of cluster</p><p>But it is still tough for the computer to select the right object with the most homogeneous part, since it can’t guess the purpose of a human. Plus, for an object with a complex pattern in itself, it is hard to get some similarity.</p><p>To summarise, the magic wand has its limits but works fine for objects which are quite homogeneous in itself but with a complex contour.</p><p>In the end, the magic wand annotation tool is not magic but pure math.</p><p>You can try the magic wand annotation tool yourself, just visit our <a href="http://www.app.moonvision.io/">MoonVision platform</a> and start a labelling job, for free.</p><p><strong>Outlook</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*5aTF2me5bGVrumjm" /></figure><p><strong>We need to go deeper! </strong>Currently we are investigating a second wand-like annotation tool which uses a more semantic constraint but pure RGB data, a deep embedding space, where you also partially define what you are interested in. So stay tuned for my next update!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a73f26acde21" width="1" height="1" alt=""><hr><p><a href="https://medium.com/moonvision/magic-wand-annotation-tool-a73f26acde21">Magic Wand Annotation Tool</a> was originally published in <a href="https://medium.com/moonvision">Moonvision</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>