ScaleFace: Uncertainty-aware Deep Metric Learning

TL;DR

ScaleFace introduces a novel deep metric learning method that estimates uncertainty directly through trainable scale values, improving face recognition and text-to-image retrieval performance with minimal additional computation.

Contribution

The paper presents ScaleFace, a new approach using trainable scale values for uncertainty estimation in deep metric learning, enhancing recognition accuracy under varying input qualities.

Findings

Outperforms existing uncertainty-aware face recognition methods.

Achieves significant improvements in text-to-image retrieval tasks.

Provides reliable uncertainty estimates with minimal computational overhead.

Abstract

The performance of modern deep learning-based systems dramatically depends on the quality of input objects. For example, face recognition quality would be lower for blurry or corrupted inputs. However, it is hard to predict the influence of input quality on the resulting accuracy in more complex scenarios. We propose an approach for deep metric learning that allows direct estimation of the uncertainty with almost no additional computational cost. The developed \textit{ScaleFace} algorithm uses trainable scale values that modify similarities in the space of embeddings. These input-dependent scale values represent a measure of confidence in the recognition result, thus allowing uncertainty estimation. We provide comprehensive experiments on face recognition tasks that show the superior performance of ScaleFace compared to other uncertainty-aware face recognition approaches. We also extend the results to the task of text-to-image retrieval showing that the proposed approach beats the competitors with significant margin.