Recall@k Surrogate Loss with Large Batches and Similarity Mixup

TL;DR

This paper introduces a differentiable surrogate loss for recall in deep image retrieval, leveraging large batch training and a similarity mixup regularization to improve performance on retrieval benchmarks.

Contribution

It proposes a novel differentiable recall surrogate loss combined with large batch training and similarity mixup regularization for enhanced deep metric learning.

Findings

Achieves state-of-the-art results on image retrieval benchmarks.

Outperforms methods using approximate average precision.

Effectively trains with very large batches despite hardware constraints.

Abstract

This work focuses on learning deep visual representation models for retrieval by exploring the interplay between a new loss function, the batch size, and a new regularization approach. Direct optimization, by gradient descent, of an evaluation metric, is not possible when it is non-differentiable, which is the case for recall in retrieval. A differentiable surrogate loss for the recall is proposed in this work. Using an implementation that sidesteps the hardware constraints of the GPU memory, the method trains with a very large batch size, which is essential for metrics computed on the entire retrieval database. It is assisted by an efficient mixup regularization approach that operates on pairwise scalar similarities and virtually increases the batch size further. The suggested method achieves state-of-the-art performance in several image retrieval benchmarks when used for deep metric learning. For instance-level recognition, the method outperforms similar approaches that train using an approximation of average precision.