<?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 Yue Bai on Medium]]></title>
        <description><![CDATA[Stories by Yue Bai on Medium]]></description>
        <link>https://medium.com/@yuebai_21410?source=rss-6329bf72a871------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*RntudXytehuJ_v3A.</url>
            <title>Stories by Yue Bai on Medium</title>
            <link>https://medium.com/@yuebai_21410?source=rss-6329bf72a871------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 17:05:45 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@yuebai_21410/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[CNN]]></title>
            <link>https://medium.com/@yuebai_21410/cnn-2b1fe1e8ed18?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b1fe1e8ed18</guid>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Fri, 20 Oct 2017 12:41:11 GMT</pubDate>
            <atom:updated>2017-10-21T09:24:02.944Z</atom:updated>
            <content:encoded><![CDATA[<h4>important knowledge</h4><p>Memory calculations</p><pre>INPUT: [224x224x3]        memory:  224*224*3=150K   weights: 0<br>CONV3-64: [224x224x64]  memory:  224*224*64=3.2M   weights: (3*3*3)*64 = 1,728<br>CONV3-64: [224x224x64]  memory:  224*224*64=3.2M   weights: (3*3*64)*64 = 36,864<br>POOL2: [112x112x64]  memory:  112*112*64=800K   weights: 0<br>CONV3-128: [112x112x128]  memory:  112*112*128=1.6M   weights: (3*3*64)*128 = 73,728<br>CONV3-128: [112x112x128]  memory:  112*112*128=1.6M   weights: (3*3*128)*128 = 147,456<br>POOL2: [56x56x128]  memory:  56*56*128=400K   weights: 0<br>CONV3-256: [56x56x256]  memory:  56*56*256=800K   weights: (3*3*128)*256 = 294,912<br>CONV3-256: [56x56x256]  memory:  56*56*256=800K   weights: (3*3*256)*256 = 589,824<br>CONV3-256: [56x56x256]  memory:  56*56*256=800K   weights: (3*3*256)*256 = 589,824<br>POOL2: [28x28x256]  memory:  28*28*256=200K   weights: 0<br>CONV3-512: [28x28x512]  memory:  28*28*512=400K   weights: (3*3*256)*512 = 1,179,648<br>CONV3-512: [28x28x512]  memory:  28*28*512=400K   weights: (3*3*512)*512 = 2,359,296<br>CONV3-512: [28x28x512]  memory:  28*28*512=400K   weights: (3*3*512)*512 = 2,359,296<br>POOL2: [14x14x512]  memory:  14*14*512=100K   weights: 0<br>CONV3-512: [14x14x512]  memory:  14*14*512=100K   weights: (3*3*512)*512 = 2,359,296<br>CONV3-512: [14x14x512]  memory:  14*14*512=100K   weights: (3*3*512)*512 = 2,359,296<br>CONV3-512: [14x14x512]  memory:  14*14*512=100K   weights: (3*3*512)*512 = 2,359,296<br>POOL2: [7x7x512]  memory:  7*7*512=25K  weights: 0<br>FC: [1x1x4096]  memory:  4096  weights: 7*7*512*4096 = 102,760,448<br>FC: [1x1x4096]  memory:  4096  weights: 4096*4096 = 16,777,216<br>FC: [1x1x1000]  memory:  1000 weights: 4096*1000 = 4,096,000<br><br>TOTAL memory: 24M * 4 bytes ~= 93MB / image (only forward! ~*2 for bwd)<br>TOTAL params: 138M parameters</pre><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b1fe1e8ed18" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Face detection and alignment]]></title>
            <link>https://medium.com/@yuebai_21410/face-detection-and-alignment-e64340dba41f?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/e64340dba41f</guid>
            <category><![CDATA[facedetection]]></category>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Wed, 11 Oct 2017 10:42:14 GMT</pubDate>
            <atom:updated>2017-10-11T11:58:10.663Z</atom:updated>
            <content:encoded><![CDATA[<h4>Comparison on different face detection and alignment methods</h4><h3>DATA SETS</h3><p>300W</p><p>AFW</p><p>PASCAL FACES</p><p>FDDB</p><p>MALF</p><p>IMAGENET</p><h3>FACE DETECTION</h3><h4>Mehthods</h4><p><a href="http://www.markusmathias.de/face_detection/#page-top">FACE DETECTION WITHOUT BELLS AND WHISTLES</a> [1]</p><p><a href="http://dlib.net/">dlib C++ library</a> (HOG+SVM)</p><h4>Implemetation</h4><ol><li><strong>Deformable Parts Model (DPM): </strong>Matlab base<em> </em><a href="http://www.robots.ox.ac.uk/~vgg/software/vgg_face/"><em>http://www.robots.ox.ac.uk/~vgg/software/vgg_face/</em></a></li><li><strong>HeadHunter: </strong>C++ and CUDA code base <a href="https://bitbucket.org/rodrigob/doppia"><em>https://bitbucket.org/rodrigob/doppia</em></a></li><li><strong>HeadHunter-Baseline</strong></li><li><strong>DLIB</strong></li></ol><h3>Comparison</h3><ol><li>DPM VS HeadHunter<strong>DPM VS HeadHunter<em> </em></strong>(pictures below reference from [1])</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/627/0*-GkE0tRnowin-4Em.png" /><figcaption>AFW</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/627/0*n5hBeIUWQHZLzR49.png" /><figcaption>PASCAL FacesFace Det</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/723/0*iQHe5BkGIZ3LAiLT.png" /><figcaption>FDDB</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*-LqazBIySJTidjSg.png" /><figcaption>MALF</figcaption></figure><p>2. HeadHunter VS Dlib on FDDB (pictures below reference from [1])</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/336/1*LY-68U1oQv9p840BieL7Xg.png" /><figcaption><a href="http://blog.csdn.net/u010333076/article/details/50637294">http://blog.csdn.net/u010333076/article/details/50637294</a></figcaption></figure><h3>Face alignment</h3><blockquote>Apparent Age Estimation From Face Images Combining General and Children-Specialized Deep Learning Models — CVPR 2016 1st [2]</blockquote><p>Face detection: DPM model</p><p>Face alignment: CLandmark[3] <a href="http://cmp.felk.cvut.cz/~uricamic/clandmark/"><strong>http://cmp.felk.cvut.cz/~uricamic/clandmark/</strong></a>ssh-keygen -t ed25519</p><blockquote>Deep expectation of real and apparent age from a single image without facial landmarks — CVPR 2015 1st [4]</blockquote><p>Face detection and alignment: DPM model</p><h3>Reference</h3><p>[1] Mathias, M., Benenson, R., Pedersoli, M., &amp; Van Gool, L. (2014, September). Face detection without bells and whistles. In <em>European Conference on Computer Vision</em> (pp. 720–735). Springer, Cham.</p><p>[2] Antipov, G., Baccouche, M., Berrani, S. A., &amp; Dugelay, J. L. (2016). Apparent age estimation from face images combining general and children-specialized deep learning models. In <em>Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops</em> (pp. 96–104).</p><p>[3] Uřičář, M., Franc, V., Thomas, D., Sugimoto, A., &amp; Hlaváč, V. (2015, May). Real-time multi-view facial landmark detector learned by the structured output SVM. In <em>Automatic Face and Gesture Recognition (FG), 2015 11th IEEE International Conference and Workshops on</em> (Vol. 2, pp. 1–8). IEEE.</p><p>[4] Rothe, R., Timofte, R., &amp; Van Gool, L. (2016). Deep expectation of real and apparent age from a single image without facial landmarks. <em>International Journal of Computer Vision</em>, 1–14.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e64340dba41f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Doppia — object_detection]]></title>
            <link>https://medium.com/@yuebai_21410/doppia-object-detection-951452c1a53f?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/951452c1a53f</guid>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Mon, 09 Oct 2017 09:28:15 GMT</pubDate>
            <atom:updated>2017-10-11T08:26:33.429Z</atom:updated>
            <content:encoded><![CDATA[<h4>problems while compiling</h4><p>Main reference <a href="https://github.com/KleinYuan/doppia#gpu-code">https://github.com/KleinYuan/doppia#gpu-code</a></p><h4>Error 1:</h4><blockquote>rodrigob-doppia-master/src/objects_detection/cascade_stages/SoftCascadeOverIntegralChannelsFastFractionalStage.cpp:24:9: error: ‘swap’ is not a member of ‘std’<br> std::swap(weak_classifier.level2_true_node, weak_classifier.level2_false_node);<br> ^</blockquote><h4>Solution 1: add #include&lt;iostream&gt; in the error file</h4><pre>#include&lt;iostream&gt;</pre><h4>Error 2:</h4><blockquote><strong>/</strong>usr<strong>/</strong>local<strong>/</strong>include<strong>/</strong>boost<strong>/</strong>variant<strong>/</strong>get.hpp:178<strong>:</strong>5<strong>:</strong> error: invalid application of ‘<strong>sizeof</strong>’ to incomplete type ‘boost<strong>::</strong>STATIC_ASSERTION_FAILURE<strong>&lt;</strong>false<strong>&gt;</strong>’<br> BOOST_STATIC_ASSERT_MSG(<br> <strong>^</strong><br>In file included from <strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>applications<strong>/</strong>objects_detection<strong>/</strong>..<strong>/</strong>..<strong>/</strong>..<strong>/</strong>libs<strong>/</strong>boost<strong>/</strong>geometry<strong>/</strong>geometries<strong>/</strong>point.hpp:26<strong>:</strong>0,<br> from <strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>applications<strong>/</strong>objects_detection<strong>/</strong>..<strong>/</strong>..<strong>/</strong>..<strong>/</strong>libs<strong>/</strong>boost<strong>/</strong>geometry<strong>/</strong>geometries<strong>/</strong>point_xy.hpp:22,<br> from <strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>objects_detection<strong>/</strong>integral_channels<strong>/</strong>IntegralChannelsForPedestrians.hpp:13,<br> from <strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>objects_detection<strong>/</strong>BaseVeryFastIntegralChannelsDetector.cpp:3<strong>:</strong><br><strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>applications<strong>/</strong>objects_detection<strong>/</strong>..<strong>/</strong>..<strong>/</strong>..<strong>/</strong>libs<strong>/</strong>boost<strong>/</strong>geometry<strong>/</strong>util<strong>/</strong>math.hpp: At global scope:<br><strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>applications<strong>/</strong>objects_detection<strong>/</strong>..<strong>/</strong>..<strong>/</strong>..<strong>/</strong>libs<strong>/</strong>boost<strong>/</strong>geometry<strong>/</strong>util<strong>/</strong>math.hpp:176<strong>:</strong>14<strong>:</strong> warning: ‘boost<strong>::</strong>geometry<strong>::</strong>math<strong>::</strong>r2d’ defined but not used [<strong>-</strong>Wunused<strong>-</strong>variable]<br> <strong>double</strong> <strong>const</strong> r2d <strong>=</strong> 1.0 <strong>/</strong> d2r;<br> <strong>^</strong><br>make[2]<strong>:</strong> <strong>***</strong> [CMakeFiles<strong>/</strong>cpp_stuff_library.dir<strong>/</strong>home<strong>/</strong>hjp<strong>/</strong>doppia<strong>/</strong>src<strong>/</strong>objects_detection<strong>/</strong>BaseVeryFastIntegralChannelsDetector.cpp.o] Error 1<br>make[1]<strong>:</strong> <strong>***</strong> [CMakeFiles<strong>/</strong>cpp_stuff_library.dir<strong>/</strong>all] Error 2<br>make[1]<strong>:</strong> <strong>***</strong> Waiting ....</blockquote><h4>Solution 2: <a href="https://bitbucket.org/rodrigob/doppia/issues/84/the-problem-while-compiling">https://bitbucket.org/rodrigob/doppia/issues/84/the-problem-while-compiling</a></h4><p>This is definitively a problem with boost. Change line 384 in common_settings.cmake from</p><pre>set(CMAKE_CXX_FLAGS &quot;${CMAKE_CXX_FLAGS} -Wall &quot;)</pre><p>to</p><pre>set(CMAKE_CXX_FLAGS &quot;${CMAKE_CXX_FLAGS} -Wall -DBOOST_VARIANT_USE_RELAXED_GET_BY_DEFAULT=1&quot;)</pre><h4>Error 3:</h4><blockquote>/rodrigob-doppia-2f93f2657960/src/objects_detection/GpuIntegralChannelsDetector.cpp:949:30: error: ‘format’ is not a member of ‘boost’</blockquote><h4>Solution 3:</h4><p><a href="https://github.com/wisk/medusa/issues/65">error: &#39;format&#39; is not a member of &#39;boost&#39; · Issue #65 · wisk/medusa</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=951452c1a53f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Something interesting about GAN]]></title>
            <link>https://medium.com/@yuebai_21410/something-interesting-about-gan-c8f44483c023?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8f44483c023</guid>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Thu, 14 Sep 2017 10:23:31 GMT</pubDate>
            <atom:updated>2017-09-14T11:08:28.267Z</atom:updated>
            <content:encoded><![CDATA[<h4>learning notes</h4><h4>Useful Link</h4><ol><li>Tutorial for GAN <a href="http://hacker.duanshishi.com/?p=1766"><em>http://hacker.duanshishi.com/?p=1766</em></a></li><li>Vedio for GAN <a href="https://www.youtube.com/watch?v=AJVyzd0rqdc"><em>https://www.youtube.com/watch?v=AJVyzd0rqdc</em></a></li></ol><h4>Problems Still</h4><ol><li>Bayesian</li><li>MCMC <a href="https://www.youtube.com/watch?v=vTUwEu53uzs"><em>https://www.youtube.com/watch?v=vTUwEu53uzs</em></a></li><li>Batch Normalization</li></ol><h4>Problem solved</h4><ol><li>Maximum likelihood estimation <a href="http://www.cnblogs.com/liliu/archive/2010/11/22/1883702.html"><em>http://www.cnblogs.com/liliu/archive/2010/11/22/1883702.html</em></a></li><li>ICA <a href="https://www.zhihu.com/question/28845451"><em>https://www.zhihu.com/question/28845451</em></a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8f44483c023" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Install Matlab 2017a & Create shortcut icons on desktop & uninstall Matlab 2017a]]></title>
            <link>https://medium.com/@yuebai_21410/install-matlab-2017a-and-create-shortcut-icons-on-desktop-944efabef6dd?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/944efabef6dd</guid>
            <category><![CDATA[matlab]]></category>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Fri, 18 Aug 2017 10:53:55 GMT</pubDate>
            <atom:updated>2017-08-18T11:25:32.341Z</atom:updated>
            <content:encoded><![CDATA[<h3>Skills for install matlab2017a</h3><p>To reactive matlab, enter the directory /usr/local/MATLAB/R2017a/bin</p><p><strong>$./activate_matlab.sh</strong></p><p>the user name can be checked by</p><p><strong>$whoami</strong></p><h3>Create icons on desktop</h3><h4>create a shortcut file</h4><p>$ sudo gedit /usr/share/applications/matlab.desktop</p><h4>write below into the file</h4><p>[Desktop Entry]<br>Name =Matlab<br>GenericName = Matlab R2017a<br>Comment = Matlab R2017a: The Language of the Techinical Computing<br>Exec = sh /usr/local/MATLAB/R2017a/bin/matlab -desktop<br>Icon = /usr/share/icons/MATLAB/matlab.png <br>StartupNotify =true<br>Terminal = false<br>Type=Application<br>Categories = Development;Matlab</p><h4>After that, the shortcut has created in Applications/Programming/ Then, modify the authority, the shortcut icon get ready to use</h4><p>sudo chmod a+x matlab.desktop</p><h4>Finally there is still some problems</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MxYiOksbyKlZVFJmAgX_vA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=944efabef6dd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Nvidia Titan X + CUDA 8.0 + cuDNN 7.0]]></title>
            <link>https://medium.com/@yuebai_21410/ubuntu16-04-nvidia-gtx-1080-cuda8-0-316a78e0f1cc?source=rss-6329bf72a871------2</link>
            <guid isPermaLink="false">https://medium.com/p/316a78e0f1cc</guid>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[ubuntu]]></category>
            <dc:creator><![CDATA[Yue Bai]]></dc:creator>
            <pubDate>Tue, 15 Aug 2017 09:10:54 GMT</pubDate>
            <atom:updated>2017-08-17T13:49:32.531Z</atom:updated>
            <content:encoded><![CDATA[<h3>Nvidia driver on ubuntu 14.04</h3><h3>Install</h3><p>Settings &gt; Repositories &gt; Additional Drivers tab &gt; select the correct nvidia driver &gt; click Apply changes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/596/1*fVSupg8rVmDeoHsdTnmW5Q.png" /></figure><h4>Verify the installation</h4><p>$ lspci –vnn | grep VGA –A 12</p><p>$ nvidia-smi</p><h3>Install Cuda 8.0</h3><p>Download toolkit package from <a href="https://developer.nvidia.com/cuda-downloads"><em>https://developer.nvidia.com/cuda-downloads</em></a></p><h4>Installation Instructions:</h4><p>$sudo dpkg -i cuda-repo-ubuntu1404–8–0-local-ga2_8.0.61–1_amd64.deb</p><p>$sudo apt-get update</p><p>$sudo apt-get install cuda</p><h4>Environment setting:</h4><p>$ sudo vi /etc/bash.bashrc</p><blockquote>export PATH=/usr/locainstallationl/cuda-8.0/bin:$PATH</blockquote><blockquote>export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH</blockquote><p>$ source /etc/bash.bashrc</p><p>$ sudo ldconfig</p><h4>Verify the environment:</h4><p>$ env</p><h4>Verify the installation:</h4><p>$ cd /usr/local/cuda-8.0/samples/1_Utilities/deviceQuery/</p><p>$ sudo make</p><p>$ ./deviceQueryInstall cuDNN 7.0</p><h3>Install cuDNN 7.0</h3><p>$ sudo cp cuda/include/cudnn.h /usr/local/cuda-8.0/include/</p><p>$ sudo cp cuda/lib64/libcudnn* /usr/local/cuda-8.0/lib64/</p><p>$ sudo chmod a+r /usr/local/cuda-8.0/include/cudnn.h</p><p>$ sudo chmod a+r /usr/local/cuda-8.0/lib64/libcudnn*</p><p>$ sudo ldconfig</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=316a78e0f1cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>