Feature Transfer Learning for Deep Face Recognition with Under-Represented Data
This addresses the issue of biased face recognition models for applications where certain subjects have limited data, though it is incremental as it builds on existing feature transfer and training techniques.
The paper tackles the problem of biased classifiers in deep face recognition due to under-represented subjects with insufficient training samples, by proposing a center-based feature transfer framework that transfers variance from regular subjects to augment the feature space of under-represented ones, achieving advantageous results on benchmarks like LFW, IJB-A, and MS-Celeb-1M compared to state-of-the-art methods.
Despite the large volume of face recognition datasets, there is a significant portion of subjects, of which the samples are insufficient and thus under-represented. Ignoring such significant portion results in insufficient training data. Training with under-represented data leads to biased classifiers in conventionally-trained deep networks. In this paper, we propose a center-based feature transfer framework to augment the feature space of under-represented subjects from the regular subjects that have sufficiently diverse samples. A Gaussian prior of the variance is assumed across all subjects and the variance from regular ones are transferred to the under-represented ones. This encourages the under-represented distribution to be closer to the regular distribution. Further, an alternating training regimen is proposed to simultaneously achieve less biased classifiers and a more discriminative feature representation. We conduct ablative study to mimic the under-represented datasets by varying the portion of under-represented classes on the MS-Celeb-1M dataset. Advantageous results on LFW, IJB-A and MS-Celeb-1M demonstrate the effectiveness of our feature transfer and training strategy, compared to both general baselines and state-of-the-art methods. Moreover, our feature transfer successfully presents smooth visual interpolation, which conducts disentanglement to preserve identity of a class while augmenting its feature space with non-identity variations such as pose and lighting.