Autoencoder based approach for the mitigation of spurious correlations

TL;DR

This paper introduces an autoencoder-based method to analyze and suppress spurious correlations in datasets, improving out-of-distribution generalization of deep neural networks, demonstrated on the GWHD 2021 dataset with a 2% ADA increase.

Contribution

The paper presents a novel autoencoder approach combined with inpainting and WBF to analyze and mitigate spurious correlations, enhancing model robustness without extensive fine-tuning.

Findings

Achieved 2% increase in Average Domain Accuracy over YOLOv5 baseline.

Effectively suppresses some spurious correlations in the GWHD 2021 dataset.

Approach is suitable for scenarios with limited model adaptation.

Abstract

Deep neural networks (DNNs) have exhibited remarkable performance across various tasks, yet their susceptibility to spurious correlations poses a significant challenge for out-of-distribution (OOD) generalization. Spurious correlations refer to erroneous associations in data that do not reflect true underlying relationships but are instead artifacts of dataset characteristics or biases. These correlations can lead DNNs to learn patterns that are not robust across diverse datasets or real-world scenarios, hampering their ability to generalize beyond training data. In this paper, we propose an autoencoder-based approach to analyze the nature of spurious correlations that exist in the Global Wheat Head Detection (GWHD) 2021 dataset. We then use inpainting followed by Weighted Boxes Fusion (WBF) to achieve a 2% increase in the Average Domain Accuracy (ADA) over the YOLOv5 baseline and consistently show that our approach has the ability to suppress some of the spurious correlations in the GWHD 2021 dataset. The key advantage of our approach is that it is more suitable in scenarios where there is limited scope to adapt or fine-tune the trained model in unseen test environments.