<p>In <a href="https://github.com/mlpack/mlpack/pull/696#discussion_r67179551">src/mlpack/methods/edge_boxes/feature_extraction_impl.hpp</a>:</p>
<pre style='color:#555'>&gt; +    // do same for dy.
&gt; +    dy.slice(i).row(0) = InImage.slice(i).row(1) - InImage.slice(i).row(0);
&gt; +    dy.slice(i).row(InImage.n_rows - 1) = InImage.slice(i).row(InImage.n_rows - 1)
&gt; +                                        - InImage.slice(i).row(InImage.n_rows - 2);
&gt; +
&gt; +    for (size_t j = 1; j &lt; InImage.n_rows-1; j++)
&gt; +      dy.slice(i).row(j) = 0.5 * ( InImage.slice(i).row(j+1) - InImage.slice(i).row(j) );
&gt; +  }  
&gt; +
&gt; +  CubeType mag(InImage.n_rows, InImage.n_cols, InImage.n_slices);
&gt; +  for (size_t i = 0; i &lt; InImage.n_slices; ++i)
&gt; +  {
&gt; +    mag.slice(i) = arma::sqrt( arma::square \
&gt; +                  ( dx.slice(i) + arma::square( dy.slice(i) ) ) );
&gt; +  }
&gt; +
</pre>
<p>Do you think we could use the sobel convolution kernel here? </p>

<p style="font-size:small;-webkit-text-size-adjust:none;color:#666;">&mdash;<br />You are receiving this because you are subscribed to this thread.<br />Reply to this email directly, <a href="https://github.com/mlpack/mlpack/pull/696/files/b38b688edfc4c2ee8bf461240f8384a70575d3a4#r67179551">view it on GitHub</a>, or <a href="https://github.com/notifications/unsubscribe/AJ4bFD3hu0xZ9nt8UMGQZUHleN04WFWqks5qMBRsgaJpZM4I1lnr">mute the thread</a>.<img alt="" height="1" src="https://github.com/notifications/beacon/AJ4bFPNLmd4HqlN4N2O2S_j0-NcbCHILks5qMBRsgaJpZM4I1lnr.gif" width="1" /></p>
<div itemscope itemtype="http://schema.org/EmailMessage">
<div itemprop="action" itemscope itemtype="http://schema.org/ViewAction">
  <link itemprop="url" href="https://github.com/mlpack/mlpack/pull/696/files/b38b688edfc4c2ee8bf461240f8384a70575d3a4#r67179551"></link>
  <meta itemprop="name" content="View Pull Request"></meta>
</div>
<meta itemprop="description" content="View this Pull Request on GitHub"></meta>
</div>