# Convolution with even-sized kernels and symmetric padding

arXiv: https://arxiv.org/abs/1903.08385 · HTML: https://mlsanity.com/paper/1903.08385/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 20 Mar 2019.

## Built on by (most cited first, 14 of 14)

- [RIC-CNN: Rotation-Invariant Coordinate Convolutional Neural Network](https://mlsanity.com/paper/2211.11812/) (2022)
- [Collapsible Linear Blocks for Super-Efficient Super Resolution](https://mlsanity.com/paper/2103.09404/) (2021)
- [PacGDC: Label-Efficient Generalizable Depth Completion with Projection Ambiguity and Consistency](https://mlsanity.com/paper/2507.07374/) (2025)
- [Context-aware Padding for Semantic Segmentation](https://mlsanity.com/paper/2109.07854/) (2021)
- [Convolutional Networks with Oriented 1D Kernels](https://mlsanity.com/paper/2309.15812/) (2023)
- [Continuous Patch Stitching for Block-wise Image Compression](https://mlsanity.com/paper/2502.16795/) (2025)
- [Padding-free Convolution based on Preservation of Differential Characteristics of Kernels](https://mlsanity.com/paper/2309.06370/) (2023)
- [PreCM: The Padding-based Rotation Equivariant Convolution Mode for Semantic Segmentation](https://mlsanity.com/paper/2411.01624/) (2024)
- [Advances in Electron Microscopy with Deep Learning](https://mlsanity.com/paper/2101.01178/) (2021)
- [DPANET:Dual Pooling Attention Network for Semantic Segmentation](https://mlsanity.com/paper/2210.05437/) (2022)
- [Improving Translation Invariance in Convolutional Neural Networks with Peripheral Prediction Padding](https://mlsanity.com/paper/2307.07725/) (2023)
- [WCCNet: Wavelet-context Cooperative Network for Efficient Multispectral Pedestrian Detection](https://mlsanity.com/paper/2308.01042/) (2023)
- [Padding Aware Neurons](https://mlsanity.com/paper/2309.08048/) (2023)
- [Enhanced Textual Feature Extraction for Visual Question Answering: A Simple Convolutional Approach](https://mlsanity.com/paper/2405.00479/) (2024)

## Abstract (arXiv)

Compact convolutional neural networks gain efficiency mainly through depthwise convolutions, expanded channels and complex topologies, which contrarily aggravate the training process. Besides, 3x3 kernels dominate the spatial representation in these models, whereas even-sized kernels (2x2, 4x4) are rarely adopted. In this work, we quantify the shift problem occurs in even-sized kernel convolutions by an information erosion hypothesis, and eliminate it by proposing symmetric padding on four sides of the feature maps (C2sp, C4sp). Symmetric padding releases the generalization capabilities of even-sized kernels at little computational cost, making them outperform 3x3 kernels in image classification and generation tasks. Moreover, C2sp obtains comparable accuracy to emerging compact models with much less memory and time consumption during training. Symmetric padding coupled with even-sized convolutions can be neatly implemented into existing frameworks, providing effective elements for architecture designs, especially on online and continual learning occasions where training efforts are emphasized.