Nonnegative Matrix Factorizations Performing Object Detection and Localization

<table>Output of the on-line detection phase on a letter envelope image presenting some handwritten digits. Test is made with LNMF after learning the USPS dataset, with <svg height="10.9125" id="M280" style="vertical-align:-0.17555pt" version="1.1" viewbox="0 0 41.512501 10.9125" width="41.512501" xmlns="http://www.w3.org/2000/svg">
<g transform="matrix(1.25,0,0,-1.25,0,10.9125)">
<g transform="translate(72,-63.27)">
<text transform="matrix(1,0,0,-1,-71.95,63.5)">
<tspan style="font-size: 12.50px; " x="0" y="0">𝑟</tspan>
<tspan style="font-size: 12.50px; " x="8.5645552" y="0">=</tspan>
<tspan style="font-size: 12.50px; " x="20.604944" y="0">8</tspan>
<tspan style="font-size: 12.50px; " x="26.856443" y="0">0</tspan>
</text>
</g>
</g>
</svg> and <svg height="13.9875" id="M281" style="vertical-align:-0.17555pt" version="1.1" viewbox="0 0 60.75 13.9875" width="60.75" xmlns="http://www.w3.org/2000/svg">
<g transform="matrix(1.25,0,0,-1.25,0,13.9875)">
<g transform="translate(72,-60.81)">
<text transform="matrix(1,0,0,-1,-71.95,61.03)">
<tspan style="font-size: 12.50px; " x="0" y="0">𝜗</tspan>
<tspan style="font-size: 12.50px; " x="10.389993" y="0">=</tspan>
<tspan style="font-size: 12.50px; " x="22.430382" y="0">2</tspan>
<tspan style="font-size: 12.50px; " x="28.681883" y="0">.</tspan>
<tspan style="font-size: 12.50px; " x="31.807632" y="0">3</tspan>
<tspan style="font-size: 12.50px; " x="38.059132" y="0">𝑒</tspan>
</text>
<text transform="matrix(1,0,0,-1,-28.32,66.03)">
<tspan style="font-size: 8.75px; " x="0" y="0">3</tspan>
</text>
</g>
</g>
</svg>.</table>

Applied Computational Intelligence and Soft Computing

fig17

Figure 17

Figure 17: Nonnegative Matrix Factorizations Performing Object Detection and Localization 