"Can a computer recognize YOUR face... even when the lights change completely?"
The short answer is yes, but not by looking at raw pixel brightness. When lighting changes, raw pixel values are completely thrown off. The Local Binary Patterns (LBP) algorithm solves this by focusing on the relationship between neighboring pixels instead. Even if the whole image gets darker or brighter, the relative difference between pixels stays the same—making the generated binary code highly robust to lighting variations.
In this video, I break down how LBP works, starting from a basic 3x3 pixel grid and scaling up to spatiotemporal analysis (LBP-TOP) for tracking facial movements over time. All the animations you see here were coded entirely in Python using the Manim library.
⏳ Timestamps:
0:00 - Intro & The Illumination Problem
1:35 - Thresholding, Neighborhoods & Uniform Patterns
4:40 - Scaling Up: Spatial Histograms
7:25 - Spatiotemporal LBP (Adding the Time Axis)
11:00 - Multi-Scale LBP & Broader Applications
💻 Source Code:This entire project is open-source. If you're curious about how these math animations were built, you can find the Manim source code, the TTS automation pipeline, and detailed technical docs on my GitHub:👉 [https://github.com/Yogurt01/lbp-manim...]
🎓 Acknowledgments:
This is an individual project developed for the Pattern Recognition course at the University of Science, VNU-HCM. A special thanks to my professors and teaching assistants for their guidance and feedback throughout the semester.
Developed by Nguyen Huu Ben. If you found this breakdown helpful, feel free to like or subscribe!
#fithcmus #patternrecognition #ai #ml #manim #computervision #machinelearning #lbp #cs #viral #easyexplain #explain #hcmus
In questa pagina del sito puoi guardare il video online [CSC14006|Pattern Recognition] How Computers Recognize Faces: Local Binary Patterns (LBP) Explained della durata di ore minuti seconda in buona qualità , che l'utente ha caricato Nguyen Huu Ben - Interesting Something! 01 gennaio 1970, condividi il link con amici e conoscenti, su youtube questo video è già stato visto 173 volte e gli è piaciuto 14 spettatori. Buona visione!