[论文解读] Eight Years of Face Recognition Research: Reproducibility, Achievements and Open Issues

Tiago de Freitas Pereira, Dominic Schmidli+4 more|arXiv (Cornell University)|Aug 8, 2022

Face recognition and analysis被引用 7

一句话总结

该论文利用可复现的开源框架，在六个数据集和五个深度学习模型上评估了八年来的面部识别进展。研究发现，当前最先进模型已能较好处理遮挡、表情和光照问题，而姿态变化超过60°、低分辨率成像以及开放集识别仍是重大挑战，尤其在低质量条件下，SCface数据集上的FNMR高达69%。

ABSTRACT

Automatic face recognition is a research area with high popularity. Many different face recognition algorithms have been proposed in the last thirty years of intensive research in the field. With the popularity of deep learning and its capability to solve a huge variety of different problems, face recognition researchers have concentrated effort on creating better models under this paradigm. From the year 2015, state-of-the-art face recognition has been rooted in deep learning models. Despite the availability of large-scale and diverse datasets for evaluating the performance of face recognition algorithms, many of the modern datasets just combine different factors that influence face recognition, such as face pose, occlusion, illumination, facial expression and image quality. When algorithms produce errors on these datasets, it is not clear which of the factors has caused this error and, hence, there is no guidance in which direction more research is required. This work is a followup from our previous works developed in 2014 and eventually published in 2016, showing the impact of various facial aspects on face recognition algorithms. By comparing the current state-of-the-art with the best systems from the past, we demonstrate that faces under strong occlusions, some types of illumination, and strong expressions are problems mastered by deep learning algorithms, whereas recognition with low-resolution images, extreme pose variations, and open-set recognition is still an open problem. To show this, we run a sequence of experiments using six different datasets and five different face recognition algorithms in an open-source and reproducible manner. We provide the source code to run all of our experiments, which is easily extensible so that utilizing your own deep network in our evaluation is just a few minutes away.

研究动机与目标

• 使用标准化的开源评估框架，评估过去八年面部识别研究的可复现性与进展。
• 分离并评估姿态、遮挡、光照、分辨率和表情等单一因素对现代基于深度学习的面部识别系统的影响。
• 基于实证评估而非仅报告结果，识别出当前已解决和仍存在的面部识别问题。
• 提供公开的源代码，使他人能够复现、扩展并将其自定义深度网络集成到评估流程中。

提出的方法

• 本研究采用开源的Bob框架，统一所有实验中的数据预处理、特征提取和评估协议。
• 在六个基准数据集中评估了五种最先进面部识别模型——ArcFace-100、ArcFace-200、Zoo-AttentionNet、Idiap-ResNet-50和FaceNet-Sandberg。
• 实验设计用于隔离特定挑战：姿态变化（使用Multi-PIE）、遮挡（AR）、光照（Multi-PIE）、低分辨率成像（GBU、SCface）以及开放集识别（IJB-C）。
• 所有模型均采用一致的图像对齐流程，且额外对ArcFace和AttentionNet等模型进行了逆向工程与对齐标准化处理。
• 评估指标包括固定FMR（0.1%）和FPIR（0.1%与100%）下的FNMR，以评估在受控与真实条件下的性能表现。
• 所有代码、超参数和实验配置均已公开发布，以确保完全可复现性与可扩展性。

实验结果

研究问题

• RQ1 在过去的八年中，现代深度学习模型在姿态、遮挡、光照或表情等方面，哪些面部识别问题已得到有效解决？
• RQ2 最先进面部识别模型在极端条件（如低分辨率图像、强姿态变化或开放集识别）下的表现如何？
• RQ3 在使用标准化、开源评估框架的前提下，现代面部识别系统性能的可复现与验证程度如何？
• RQ4 特定损失函数（如ArcFace）在提升识别鲁棒性方面，对各类挑战因素起到了多大作用？

主要发现

• ArcFace-100与Zoo-AttentionNet在AR数据集上对遮挡情况实现了0%的FNMR（FMR为0.1%），相比以往系统有显著提升。
• 在手机端数据集（MOBIO）上，面部识别性能大幅提升，表情变化导致的FNMR从44%降至0%。
• 姿态变化在±60°以内时处理良好，但在此范围之外性能急剧下降，尤其在Multi-PIE数据集上表现明显。
• 低分辨率和低质量的查询图像导致极高的失败率：ArcFace-100在SCface数据集‘far’协议下FNMR高达69%，表明这是当前重大开放挑战。
• 开放集识别仍是显著问题，IJB-C数据集上识别率从90%（FPIR=100%）骤降至仅35%（FPIR=0.1%），凸显理想化与真实世界性能之间的差距。
• 尽管取得进展，可复现性仍存在问题，部分关键模型（如ArcFace）的预训练权重已被移除或对齐流程发生变更，影响了结果的一致性。