A Sobel-Gradient MLP Baseline for Handwritten Character Recognition

Azam Nouri
8/16/2025
cs.CVcs.LG

No official code found. Best implementation: Meetra21/Gradient-Driven-MLP (1 star, Apache-2.0, last commit 2025-10-14).

Abstract

We revisit the classical Sobel operator to ask a simple question: Are first-order edge maps sufficient to drive an all-dense multilayer perceptron (MLP) for handwritten character recognition (HCR), as an alternative to convolutional neural networks (CNNs)? Using only horizontal and vertical Sobel derivatives as input, we train an MLP on MNIST and EMNIST Letters. Despite its extreme simplicity, the resulting network reaches 98% accuracy on MNIST digits and 92% on EMNIST letters -- approaching CNNs while offering a smaller memory footprint and transparent features. Our findings highlight that much of the class-discriminative information in handwritten character images is already captured by first-order gradients, making edge-aware MLPs a compelling option for HCR.

View on arXivView PDF

Code Implementations(1)

Meetra21/Gradient-Driven-MLPCommunity63%
Apache-2.0

Code and resources for "A Sobel-Gradient MLP Baseline for Handwritten Character Recognition"

10Jupyter NotebookAug 11, 20251 years agoApache-2.0

Similar Papers

An MLP Baseline for Handwriting Recognition Using Planar Curvature and Gradient Orientation

88% match2025

Rethinking Deep Learning: Propagating Information in Neural Networks without Backpropagation and Statistical Optimization

79% match2024

Evaluating Machine Learning Approaches for ASCII Art Generation

79% match1 repo32025

Multichannel Attention Networks with Ensembled Transfer Learning to Recognize Bangla Handwritten Charecter

78% match2024

Revisiting N-Gram Models: Their Impact in Modern Neural Networks for Handwritten Text Recognition

78% match2024

Efficient Multi-domain Text Recognition Deep Neural Network Parameterization with Residual Adapters

78% match1 repo42024

Cite this paper

@article{nouri2025sobelgradient,
  title  = {A Sobel-Gradient MLP Baseline for Handwritten Character Recognition},
  author = {Azam Nouri},
  year   = {2025},
  eprint = {2508.11902},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2508.11902v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2508.11902.svg)](https://codeofpaper.com/paper/2508.11902)

Discussion