Fully Convolutional Fractional Scaling

TL;DR

This paper presents FCFS, a fully convolutional fractional scaling component that enables non-integer scaling in convolutional networks, with a simple architecture and practical implementations.

Contribution

Introduction of FCFS, a novel fully convolutional fractional scaling module supporting non-integer scaling with an efficient single-layer design.

Findings

Supports non-integer scaling in convolutional networks

Provides practical code implementations for common scaling methods

Enables flexible input size processing in fully convolutional architectures

Abstract

We introduce a fully convolutional fractional scaling component, FCFS. Fully convolutional networks can be applied to any size input and previously did not support non-integer scaling. Our architecture is simple with an efficient single layer implementation. Examples and code implementations of three common scaling methods are published.