<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Sergeiy Varyukhin</title>
    <description>The latest articles on DEV Community by Sergeiy Varyukhin (@sergeiyvaryukhin).</description>
    <link>https://dev.to/sergeiyvaryukhin</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4242620%2F6b7b8f68-57a8-462f-b081-53e6b79051e7.png</url>
      <title>DEV Community: Sergeiy Varyukhin</title>
      <link>https://dev.to/sergeiyvaryukhin</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/sergeiyvaryukhin"/>
    <language>en</language>
    <item>
      <title>I built haralickt: &gt; 20x faster GPU-accelerated Haralick texture features (GLCM) in PyTorch with identical precision to Mahotas</title>
      <dc:creator>Sergeiy Varyukhin</dc:creator>
      <pubDate>Fri, 13 Nov 2026 12:14:25 +0000</pubDate>
      <link>https://dev.to/sergeiyvaryukhin/i-built-haralickt-20x-faster-gpu-accelerated-haralick-texture-features-glcm-in-pytorch-with-12ln</link>
      <guid>https://dev.to/sergeiyvaryukhin/i-built-haralickt-20x-faster-gpu-accelerated-haralick-texture-features-glcm-in-pytorch-with-12ln</guid>
      <description>&lt;p&gt;Hi everyone! &lt;br&gt;
If you work with texture analysis (medical imaging, satellite data, or industrial defect detection), you've likely noticed that classic libraries like mahotas and scikit-image compute Gray-Level Co-occurrence Matrices (GLCM) and Haralick features strictly on the CPU. On large datasets, this quickly becomes a major pipeline bottleneck.&lt;br&gt;
To solve this, I wrote and open-sourced haralickt — a library that offloads these computations to the GPU using vectorized matrix operations in PyTorch and CUDA.&lt;br&gt;
Key highlights:&lt;br&gt;
• &amp;gt; 20x faster than traditional CPU implementations on high-resolution images and batches.&lt;br&gt;
• Identical numerical precision to classic implementations (verified against mahotas).&lt;br&gt;
• PyTorch-native: works seamlessly with torch.Tensor on GPU without costly CPU-GPU data transfers.&lt;br&gt;
• Open Source (MPL 2.0): free to use in research and commercial projects.&lt;br&gt;
Quick installation:&lt;br&gt;
bash&lt;br&gt;
pip install haralickt&lt;br&gt;
Используйте код с осторожностью.&lt;br&gt;
GitHub repository &amp;amp; benchmarks: [&lt;a href="https://github.com/SergeiyVaryukhin/haralickt-1.0.0" rel="noopener noreferrer"&gt;https://github.com/SergeiyVaryukhin/haralickt-1.0.0&lt;/a&gt;]&lt;br&gt;
This is my first public release (v1.0.0), so I would really appreciate any feedback, bug reports, or ideas for new features. If you find it useful for your projects, a star on GitHub would mean a lot!&lt;/p&gt;

</description>
      <category>opensource</category>
      <category>performance</category>
      <category>python</category>
      <category>showdev</category>
    </item>
  </channel>
</rss>
