A Light CNN for Deep Face Representation with Noisy Labels
About
The volume of convolutional neural network (CNN) models proposed for face recognition has been continuously growing larger to better fit large amount of training data. When training data are obtained from internet, the labels are likely to be ambiguous and inaccurate. This paper presents a Light CNN framework to learn a compact embedding on the large-scale face data with massive noisy labels. First, we introduce a variation of maxout activation, called Max-Feature-Map (MFM), into each convolutional layer of CNN. Different from maxout activation that uses many feature maps to linearly approximate an arbitrary convex activation function, MFM does so via a competitive relationship. MFM can not only separate noisy and informative signals but also play the role of feature selection between two feature maps. Second, three networks are carefully designed to obtain better performance meanwhile reducing the number of parameters and computational costs. Lastly, a semantic bootstrapping method is proposed to make the prediction of the networks more consistent with noisy labels. Experimental results show that the proposed framework can utilize large-scale noisy data to learn a Light model that is efficient in computational costs and storage spaces. The learned single network with a 256-D representation achieves state-of-the-art results on various face benchmarks without fine-tuning. The code is released on https://github.com/AlfredXiangWu/LightCNN.
Key insights
How does the paper address noisy labels in face recognition?
It introduces a Light CNN using Max-Feature-Map (MFM) activation, a maxout variant that suppresses noisy signals via competition between feature maps, plus a semantic bootstrapping method to relabel noisy training data.
What distinguishes this work from prior face recognition methods?
Unlike ReLU-based CNNs, MFM uses parameter-free competitive inhibition for better generalization. Light CNN-29 achieves 99.33% LFW accuracy with 12.6M parameters, roughly 10x smaller and 5x faster than VGG, evaluated across five benchmark types.
What are the practical uses of the Light CNN framework?
The authors state the framework has potential value for real-time face recognition systems and demonstrate deployment on MaPU embedded hardware with 40ms feature extraction, suggesting suitability for embedding devices and smartphones.
What limitations and future work do the authors note?
On Celebrity-1000, Light CNN-29 underperforms GoogleNet+NAN because models were not trained on that dataset. The authors note performance gains diminish after two bootstrapping rounds as identity count drops from 99,891 to 79,077.