TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Neural Network Recipe: Recognize Handwritten Digits with 95% Accuracy-C++/nolibs

3 pointsby Atrix256about 8 years ago

1 comment

Atrix256about 8 years ago
If you tried the demo and were remarkably unimpressed, give it a try again and it ought to work better (make sure and manually refresh - F5 so you don&#x27;t have the old page cached).<p>I read how the MNIST data was generated and followed suit: Find the bounding box of the drawn image and scale it down to 20x20, preserving the aspect ratio. Put that 20x20 image into a 28x28 image using the center of mass of the pixels as the center point.<p>There is still some error due to how different mouse drawn and hand drawn numbers look, but I&#x27;ve found it to be a lot more accurate.<p>Demo: <a href="http:&#x2F;&#x2F;demofox.org&#x2F;MNIST1.html" rel="nofollow">http:&#x2F;&#x2F;demofox.org&#x2F;MNIST1.html</a><p>Example training data to see what i mean about difference between hand drawn and mouse drawn images: <a href="https:&#x2F;&#x2F;i2.wp.com&#x2F;blog.demofox.org&#x2F;wp-content&#x2F;uploads&#x2F;2017&#x2F;03&#x2F;mnist_100_digits1.png" rel="nofollow">https:&#x2F;&#x2F;i2.wp.com&#x2F;blog.demofox.org&#x2F;wp-content&#x2F;uploads&#x2F;2017&#x2F;0...</a><p>Thanks for the feedback on that... it was a mistake rolling it out in such a broken state.