Exploring the ability of CNNs to generalise to previously unseen scales over wide scale ranges

04/03/2020
by   Ylva Jansson, et al.
0

The ability to handle large scale variations is crucial for many real world visual tasks. A straightforward approach for handling scale in a deep network is to process an image at several scales simultaneously in a set of scale channels. Scale invariance can then, in principle, be achieved by using weight sharing between the scale channels together with max or average pooling over the outputs from the scale channels. The ability of such scale channel networks to generalise to scales not present in the training set over significant scale ranges has, however, not previously been explored. We, therefore, present a theoretical analysis of invariance and covariance properties of scale channel networks and perform an experimental evaluation of the ability of different types of scale channel networks to generalise to previously unseen scales. We identify limitations of previous approaches and propose a new type of foveated scale channel architecture, where the scale channels process increasingly larger parts of the image with decreasing resolution. Our proposed FovMax and FovAvg networks perform almost identically over a scale range of 8 also when training on single scale training data and give improvements in the small sample regime.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/11/2021

Scale-invariant scale-channel networks: Deep networks that generalise to previously unseen scales

The ability to handle large scale variations is crucial for many real wo...
research
05/08/2023

Riesz networks: scale invariant neural networks in a single forward pass

Scale invariance of an algorithm refers to its ability to treat objects ...
research
11/30/2020

Scale-covariant and scale-invariant Gaussian derivative networks

This article presents a hybrid approach between scale-space theory and d...
research
02/05/2023

Achieving Robust Generalization for Wireless Channel Estimation Neural Networks by Designed Training Data

In this paper, we propose a method to design the training data that can ...
research
09/27/2021

Learning from Small Samples: Transformation-Invariant SVMs with Composition and Locality at Multiple Scales

Motivated by the problem of learning when the number of training samples...
research
04/05/2022

Arbitrary-Scale Image Synthesis

Positional encodings have enabled recent works to train a single adversa...
research
08/27/2020

Wi-Fi All-Channel Analyzer

In this paper, we present WACA, the first system to simultaneously measu...

Please sign up or login with your details

Forgot password? Click here to reset