On Robustness and Transferability of Convolutional Neural Networks

by   Josip Djolonga, et al.

Modern deep convolutional networks (CNNs) are often criticized for not generalizing under distributional shifts. However, several recent breakthroughs in transfer learning suggest that these networks can cope with severe distribution shifts and successfully adapt to new tasks from a few training examples. In this work we revisit the out-of-distribution and transfer performance of modern image classification CNNs and investigate the impact of the pre-training data size, the model scale, and the data preprocessing pipeline. We find that increasing both the training set and model sizes significantly improve the distributional shift robustness. Furthermore, we show that, perhaps surprisingly, simple changes in the preprocessing such as modifying the image resolution can significantly mitigate robustness issues in some cases. Finally, we outline the shortcomings of existing robustness evaluation datasets and introduce a synthetic dataset we use for a systematic analysis across common factors of variation.


page 5

page 7

page 15

page 16

page 18

page 20

page 21

page 22


Measuring Robustness to Natural Distribution Shifts in Image Classification

We study how robust current ImageNet models are to distribution shifts a...

Using transfer learning to detect galaxy mergers

We investigate the use of deep convolutional neural networks (deep CNNs)...

Powerset Convolutional Neural Networks

We present a novel class of convolutional neural networks (CNNs) for set...

CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters

Currently, many theoretical as well as practically relevant questions to...

AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty

Modern deep neural networks can achieve high accuracy when the training ...

A Method for Restoring the Training Set Distribution in an Image Classifier

Convolutional Neural Networks are a well-known staple of modern image cl...

A Data Driven Approach for Compound Figure Separation Using Convolutional Neural Networks

A key problem in automatic analysis and understanding of scientific pape...

Code Repositories