Batch-Instance Normalization for Adaptively Style-Invariant Neural Networks

TL;DR

This paper introduces Batch-Instance Normalization (BIN), a simple yet effective method to explicitly normalize style variations in images, enhancing neural network robustness across different visual recognition tasks.

Contribution

BIN is a novel normalization technique that selectively removes style information, improving recognition performance and adaptability in various visual tasks.

Findings

BIN improves recognition accuracy across multiple scenarios.

BIN effectively adapts to different tasks like classification and style transfer.

BIN is easy to implement with minimal code.

Abstract

Real-world image recognition is often challenged by the variability of visual styles including object textures, lighting conditions, filter effects, etc. Although these variations have been deemed to be implicitly handled by more training data and deeper networks, recent advances in image style transfer suggest that it is also possible to explicitly manipulate the style information. Extending this idea to general visual recognition problems, we present Batch-Instance Normalization (BIN) to explicitly normalize unnecessary styles from images. Considering certain style features play an essential role in discriminative tasks, BIN learns to selectively normalize only disturbing styles while preserving useful styles. The proposed normalization module is easily incorporated into existing network architectures such as Residual Networks, and surprisingly improves the recognition performance in various scenarios. Furthermore, experiments verify that BIN effectively adapts to completely different tasks like object classification and style transfer, by controlling the trade-off between preserving and removing style variations. BIN can be implemented with only a few lines of code using popular deep learning frameworks.