Deep Nets with Subsampling Layers Unwittingly Discard Useful Activations at Test-Time

TL;DR

This paper challenges the common practice of discarding activations in subsampling layers of deep nets, proposing a method to utilize these discarded features at test time to enhance prediction accuracy across various tasks.

Contribution

The authors introduce a search and aggregate method to leverage discarded activations in subsampling layers, improving test-time performance in image classification and segmentation.

Findings

Consistent performance improvements across nine architectures.

Enhancement of existing test-time augmentation techniques.

Applicable to multiple datasets and tasks.

Abstract

Subsampling layers play a crucial role in deep nets by discarding a portion of an activation map to reduce its spatial dimensions. This encourages the deep net to learn higher-level representations. Contrary to this motivation, we hypothesize that the discarded activations are useful and can be incorporated on the fly to improve models' prediction. To validate our hypothesis, we propose a search and aggregate method to find useful activation maps to be used at test time. We applied our approach to the task of image classification and semantic segmentation. Extensive experiments over nine different architectures on multiple datasets show that our method consistently improves model test-time performance, complementing existing test-time augmentation techniques. Our code is available at https://github.com/ca-joe-yang/discard-in-subsampling.