Exploring Factors for Improving Low Resolution Face Recognition

Abstract

State-of-the-art deep face recognition approaches report near perfect performance on popular benchmarks, e.g., Labeled Faces in the Wild. However, their performance deteriorates significantly when they are applied on low quality images, such as those acquired by surveillance cameras. A further challenge for low resolution face recognition for surveillance applications is the matching of recorded low resolution probe face images with high resolution reference images, which could be the case in watchlist scenarios. In this paper, we have addressed these problems and investigated the factors that would contribute to the identification performance of the state-of-the-art deep face recognition models when they are applied to low resolution face recognition under mismatched conditions. We have observed that the following factors affect performance in a positive way: appearance variety and resolution distribution of the training dataset, resolution matching between the gallery and probe images, and the amount of information included in the probe images. By leveraging this information, we have utilized deep face models trained on MS-Celeb-1M and finetuned on VGGFace2 dataset and achieved state-of-the-art accuracies on the SCFace and ICB-RW benchmarks, even without using any training data from the datasets of these benchmarks.

CNN-based Gender Prediction in Uncontrolled Environments

“…Convolutional neural networks in many cases give better results than traditional networks [27][28]. However, we see that the performance is significantly reduced when these networks work with low-resolution images [29]. We briefly explain how CNN layer structure works.…”

Deep Rival Penalized Competitive Learning for Low-resolution Face Recognition

“…Since massive wild or surveillance LR faces are not available for training, the existing LR methods are based on down-sampling HR faces. [27] investigated the face crop factors that would affect the LR face recognition performance, and trained the model by leveraging the factors to improve the accuracy. [28] conducted experimental research in actual surveillance applications, and empirically evaluated super-resolution methods for LR face recognition.…”

Design and Analysis of an Expert System for the Detection and Recognition of Criminal Faces

“…Research is still making progress in the direction of developing an intuitive and intelligent system that is comparable to human perception [3,4]. Over the course of history, several distinct algorithms and approaches to the identifcation of faces have been devised [5][6][7][8][9]. Tese techniques are centered on tracing the contours of the face and isolating its various characteristics, such as the eyes, nose, and mouth.…”