Understanding Image Normalization in CNNs

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Laurent Perrinet
Format: Recurso digital
Sprache:Englisch
Veröffentlicht: Zenodo 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866901521415798784
author Laurent Perrinet
author_facet Laurent Perrinet
contents <p>[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.14602370.svg)](https://doi.org/10.5281/zenodo.14602370)</p> <p>#ConvolutionalNeuralNetworks (#CNNs in short) are immensely useful for many #imageProcessing tasks and much more... Yet you sometimes encounter some bits of code </p> <p>Have you ever wondered about the origins of the values for image normalization in #imagenet ?</p> <p><br>* Mean: `[0.485, 0.456, 0.406]` (for R, G and B channels respectively)<br>* Std: `[0.229, 0.224, 0.225]`</p> <p>Strangest to me is the need for a three-digits precision. Here,  after finding the origin of these numbers for MNIST and ImageNet, I am testing if that precision is really important : guess what, it is not !</p> <p>* https://laurentperrinet.github.io/sciblog/posts/2024-12-09-normalizing-images-in-convolutional-neural-networks.html<br>* https://nbviewer.org/github/laurentperrinet/2024-12-09-normalizing-images-in-convolutional-neural-networks/blob/main/2024-12-09-normalizing-images-in-convolutional-neural-networks.ipynb</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_14602370
institution Zenodo
language eng
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Understanding Image Normalization in CNNs
Laurent Perrinet
Deep Learning
CNN
normalization
Image processing
Computer vision
<p>[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.14602370.svg)](https://doi.org/10.5281/zenodo.14602370)</p> <p>#ConvolutionalNeuralNetworks (#CNNs in short) are immensely useful for many #imageProcessing tasks and much more... Yet you sometimes encounter some bits of code </p> <p>Have you ever wondered about the origins of the values for image normalization in #imagenet ?</p> <p><br>* Mean: `[0.485, 0.456, 0.406]` (for R, G and B channels respectively)<br>* Std: `[0.229, 0.224, 0.225]`</p> <p>Strangest to me is the need for a three-digits precision. Here,  after finding the origin of these numbers for MNIST and ImageNet, I am testing if that precision is really important : guess what, it is not !</p> <p>* https://laurentperrinet.github.io/sciblog/posts/2024-12-09-normalizing-images-in-convolutional-neural-networks.html<br>* https://nbviewer.org/github/laurentperrinet/2024-12-09-normalizing-images-in-convolutional-neural-networks/blob/main/2024-12-09-normalizing-images-in-convolutional-neural-networks.ipynb</p>
title Understanding Image Normalization in CNNs
topic Deep Learning
CNN
normalization
Image processing
Computer vision
url https://doi.org/10.5281/zenodo.14602370