Efficient Adaptive Ensembling for Image Classification

TL;DR

This paper introduces an efficient adaptive ensembling method that improves image classification accuracy without increasing model complexity, by fine-tuning a combination layer on two trained EfficientNet-b0 models.

Contribution

The paper presents a novel ensembling approach that is computationally efficient and enhances accuracy, addressing the complexity issues of traditional ensembling methods.

Findings

Outperforms state-of-the-art by 0.5% accuracy on average

Reduces parameters by 5-60 times compared to previous methods

Lowers FLOPS by 10-100 times across benchmark datasets

Abstract

In recent times, with the exception of sporadic cases, the trend in Computer Vision is to achieve minor improvements compared to considerable increases in complexity. To reverse this trend, we propose a novel method to boost image classification performances without increasing complexity. To this end, we revisited ensembling, a powerful approach, often not used properly due to its more complex nature and the training time, so as to make it feasible through a specific design choice. First, we trained two EfficientNet-b0 end-to-end models (known to be the architecture with the best overall accuracy/complexity trade-off for image classification) on disjoint subsets of data (i.e. bagging). Then, we made an efficient adaptive ensemble by performing fine-tuning of a trainable combination layer. In this way, we were able to outperform the state-of-the-art by an average of 0.5$\%$ on the accuracy, with restrained complexity both in terms of the number of parameters (by 5-60 times), and the FLoating point Operations Per Second (FLOPS) by 10-100 times on several major benchmark datasets.