Learning Compact Face Representation

Abstract

This paper addresses the problem of producing very compact representation of a face image for large-scale face search and analysis tasks. In tradition, the compactness of face representation is achieved by a dimension reduction step after representation extraction. However, the dimension reduction usually degrades the discriminative ability of the original representation drastically. In this paper, we present a deep learning framework which optimizes the compactness and discriminative ability jointly. The learnt representation can be as compact as 32 bit (same as the int32) and still produce highly discriminative performance (91.4% on LFW benchmark). Based on the extreme compactness, we show that traditional face analysis tasks (e.g. gender analysis) can be effectively solved by a Look-Up- Table approach given a large-scale face data set.

Labeled Faces in the Wild: A Survey

“…In the first paper from the Face++ group, a new structure, which they term the pyramid CNN [34] is used. It conducts supervised training of a deep neural network one layer at a time, thus greatly reducing computation.…”

The effect of pets on happiness: A data-driven approach via large-scale social media

“…Another area of the related work is applying deep learning techniques to solving user-content related computer vision tasks: pet detection and face analysis. Deep neural networks such as CNN have shown great success in the tasks of object recognition [7], [9], as well as face analysis including face detection and demographics inference [3], [14]. We build the pet classifier based on CNN which is further combined with timeline analysis for pet owner identification, and apply the state-of-the-art face++ engine for face analysis for a fine-grained analysis on the pet effect on happiness.…”

“…We use Face++ for selfie image detection and demographics inference since Face++ is the state-of-the-art openaccess face engine on tasks of face detection and face analysis [3], [14]. We first use selfie detection to target users who have posted multiple selfies in a certain time frame and then apply face analysis on all the images we collected from these users within six months.…”

Spatio-Temporal Wardrobe Generation of Actors’ Clothing in Video Content

“…Betaface [22] and openbiometrics [23]). Since improving face detection algorithms is out of scope for this paper, we make use of the commercial Face++ algorithm [24]. An evaluation is done by first comparing the output of several commercial detectors, and additionally evaluating the estimated face coordinates.…”