A Deep Face Identification Network Enhanced by Facial Attributes Prediction

Abstract

In this paper, we propose a new deep framework which predicts facial attributes and leverage it as a soft modality to improve face identification performance. Our model is an end to end framework which consists of a convolutional neural network (CNN) whose output is fanned out into two separate branches; the first branch predicts facial attributes while the second branch identifies face images. Contrary to the existing multi-task methods which only use a shared CNN feature space to train these two tasks jointly, we fuse the predicted attributes with the features from the face modality in order to improve the face identification performance. Experimental results show that our model brings benefits to both face identification as well as facial attribute prediction performance, especially in the case of identity facial attributes such as gender prediction. We tested our model on two standard datasets annotated by identities and face attributes. Experimental results indicate that the proposed model outperforms most of the current existing face identification and attribute prediction methods.

Effective De-identification Generative Adversarial Network for Face Anonymization

“…Traditional face DeId methods, like blurring or pixelation, are commonly used to prevent observers from recognizing the original identity in an image [30,40]. By analyzing the performance of face identity recognizer, Oh et al [36] showed that the commonly used DeId methods are insecure, especially that the recent deep recognizers can achieve high face re-identification rates [9,33,49]. Alternatively, Newton et al [35] proposed a k-Same algorithm to synthesize de-identified face images for k-anonymized privacy protection [48].…”

A Self Supervised Defending Mechanism Against Adversarial Iris Attacks based on Wavelet Transform

“…Recently, deep learning models have performed tremendously in a large range of applications like biometrics [10,11], security [12,13], autonomous vehicle control systems and Spam Filtering. However these models are more susceptible to manipulated input data which is called adversarial examples.…”

A Survey of Deep Facial Attribute Analysis

“…In addition, facial attributes have been taken as auxiliary and complementary information for many facerelated tasks, such as face recognition [57,91,112], face detection [86], and facial landmark localization [139]. Kumar et al [57] first introduce the concept of 'attribute' to facilitate face verification by compact visual descriptions and low-level attribute features.…”