Boosting Semi-Supervised Face Recognition with Noise Robustness

TL;DR

This paper introduces a noise-robust semi-supervised face recognition method called NRoLL, which effectively handles label errors during auto-labelling, achieving high accuracy with limited labelled data.

Contribution

The paper proposes GroupNet for identifying and preserving clean labels, and develops NRoLL, a semi-supervised approach that boosts face recognition performance under noisy label conditions.

Findings

GroupNet achieves leading accuracy with over 50% noisy labels.

NRoLL effectively leverages limited labelled data for semi-supervised learning.

Our method outperforms state-of-the-art approaches on multiple benchmarks.

Abstract

Although deep face recognition benefits significantly from large-scale training data, a current bottleneck is the labelling cost. A feasible solution to this problem is semi-supervised learning, exploiting a small portion of labelled data and large amounts of unlabelled data. The major challenge, however, is the accumulated label errors through auto-labelling, compromising the training. This paper presents an effective solution to semi-supervised face recognition that is robust to the label noise aroused by the auto-labelling. Specifically, we introduce a multi-agent method, named GroupNet (GN), to endow our solution with the ability to identify the wrongly labelled samples and preserve the clean samples. We show that GN alone achieves the leading accuracy in traditional supervised face recognition even when the noisy labels take over 50\% of the training data. Further, we develop a semi-supervised face recognition solution, named Noise Robust Learning-Labelling (NRoLL), which is based on the robust training ability empowered by GN. It starts with a small amount of labelled data and consequently conducts high-confidence labelling on a large amount of unlabelled data to boost further training. The more data is labelled by NRoLL, the higher confidence is with the label in the dataset. To evaluate the competitiveness of our method, we run NRoLL with a rough condition that only one-fifth of the labelled MSCeleb is available and the rest is used as unlabelled data. On a wide range of benchmarks, our method compares favorably against the state-of-the-art methods.