Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not?

By Erjin Zhou, Zhimin Cao, Qi Yin. 2015

198 citations recorded by the source corpus.

Face recognition performance improves rapidly with the recent deep learning technique developing and underlying large training dataset accumulating. In this paper, we report our observations on how big data impacts the recognition performance. According to these observations, we build our Megvii Face Recognition System, which achieves 99.50% accuracy on the LFW benchmark, outperforming the previous state-of-the-art. Furthermore, we report the performance in a real-world security certification scenario. There still exists a clear gap between machine recognition and human performance. We summarize our experiments and present three challenges lying ahead in recent face recognition. And we indicate several possible solutions towards these challenges. We hope our work will stimulate the community's discussion of the difference between research benchmark and real-world applications.

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Evaluation & Benchmarks

Citations in the corpus

Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition (2016, methodology): “In this section we report on experiments for evaluating the eﬃcacy of this approach in fooling DNN B and DNN C under semi-controlled imaging conditions.”
Defeating Image Obfuscation with Deep Learning (2016, methodology): “Zhou et al. [56] and Parkhi et al. [36] presented CNNs for the Labeled Faces in the Wild database [17] supplemented with large datasets of Web images.”
Person Recognition in Personal Photo Collections (2015, background): “Both settings have seen a recent shift from sophisticated classifiers based on hand-crafted features and metric learning approaches [20, 7, 5, 30, 27, 42, 1], towards methods based on deep learning [38, 37, 44, 34, 28, 39, 21].”
Robust Face Recognition via Multimodal Deep Face Representation (2015, background): “Zhou et al. [18] investigated the inﬂuence of distribution and size of training data to the performance of CNN.”
Deep Face Recognition: A Survey (2018, methodology): “Recently, data augmentation has been widely used in deep FR algorithms [99], [228], [46], [177], [145], [146], [150], [160].”
Person Recognition in Personal Photo Collections (2017, background): “The benchmark has nearly saturated in the recent years, attributing to the deep features [3], [5], [6], [7], [8], [9], [10] trained on large scale face databases that outperform the traditional methods involving sophisticated classiﬁers based on hand-crafted features and metric learning approaches…”
Targeting Ultimate Accuracy: Face Recognition via Deep Embedding (2015, methodology): “Many smart methods have been used in the first step, such as joint learning [6, 8, 9], multistage feature and supervision [6, 7, 9], multi-patch features [2, 6, 7, 8, 9] and sophisticated network structure [12].”
Sketch Me That Shoe (2016, background): “It is increasingly clear that CNN performance ceiling in practice is imposed by limits on available data, with additional data improving performance [33].”
Range Loss for Deep Face Recognition with Long-Tailed Training Data (2016, unknown): “Empirical studies and analysis show that classes with more samples will pose greater impact on the feature learning procedure [37, 21] and inversely cripple the models ability on tailed part.”
Twitter's Glass Ceiling: The Effect of Perceived Gender on Online Visibility (2016, background): “Online services such as http://how-old.net by Microsoft (Team 2015) and Face++ have begun to provide APIs for detecting gender, race, age, etc. from a profile image and achieves high levels of recognition accuracy (Zhou, Cao, and Yin 2015).”
Learning Pose-Aware Models for Pose-Invariant Face Recognition in the Wild (2019, background): “and varied data for training the CNN in [7], [29], [30].”
When Face Recognition Meets with Deep Learning: An Evaluation of Convolutional Neural Networks for Face Recognition (2015, background): “As a result, the CNN architectures used for face recognition [24, 21, 25, 33, 28] are usually different from those for object recognition [14, 20, 26, 9].”