A Sobel-Gradient MLP Baseline for Handwritten Character Recognition

TL;DR

This paper demonstrates that a simple MLP using only Sobel edge maps can achieve near state-of-the-art accuracy in handwritten character recognition, offering a lightweight and transparent alternative to CNNs.

Contribution

It introduces a minimalistic edge-based input approach for MLPs in handwritten character recognition, challenging the necessity of complex CNN architectures.

Findings

Achieves 98% accuracy on MNIST with Sobel-MLP

Achieves 92% accuracy on EMNIST Letters with Sobel-MLP

Edge features capture most class-discriminative information

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.