CVJul 10, 2017

Checkerboard artifact free sub-pixel convolution: A note on sub-pixel convolution, resize convolution and convolution resize

arXiv:1707.02937v1195 citations
Originality Incremental advance
AI Analysis

This addresses a specific artifact problem in image generation and dense labeling tasks, but it is incremental as it builds on existing methods like sub-pixel and resize convolution.

The paper tackled checkerboard artifacts in deconvolution layers by proposing a new initialization method called convolution NN resize for sub-pixel convolution, which eliminates artifacts immediately after initialization and achieves smaller test errors with the same computational complexity as resize convolution.

The most prominent problem associated with the deconvolution layer is the presence of checkerboard artifacts in output images and dense labels. To combat this problem, smoothness constraints, post processing and different architecture designs have been proposed. Odena et al. highlight three sources of checkerboard artifacts: deconvolution overlap, random initialization and loss functions. In this note, we proposed an initialization method for sub-pixel convolution known as convolution NN resize. Compared to sub-pixel convolution initialized with schemes designed for standard convolution kernels, it is free from checkerboard artifacts immediately after initialization. Compared to resize convolution, at the same computational complexity, it has more modelling power and converges to solutions with smaller test errors.

Code Implementations3 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes