Abstract:In real-world applications, face recognition, and person re-identification are subject to image degradations such as motion blur, atmospheric turbulence, or upsampling artifacts—which are known to lower performance. This work directly addresses challenges in low-quality scenarios with 1) practical, novel updates to training and inference, which improve robustness to realistic distortions in face recognition and person re-identification and 2) new datasets for long-distance recognition. We propose a method that progressively learns from images prone to soft and strong distortions caused mainly by atmospheric turbulence. The method has a novel distortion loss to improve robustness, which is empirically shown to be highly effective in low-quality scenarios. Two further strategies are proposed to integrate distortion augmentation while also retaining the highest performance in high-quality scenarios. First, during training, an adaptive weighting schedule, which leverages the construction of different levels of distortion augmentation, is used to train the model in an easy-to-hard manner. The second, at inference, is a magnitude-weighted fusion of features from the parallel models used to retain the highest robustness across both high-quality and low-quality imagery. Different from prior work, our model does not leverage any image restoration or style transfer technique, and we are the first to employ explicit distortion weighting during training and evaluation. Our model achieves the best performances compared to prior works on face recognition and person re-identification benchmarks, including IARPA Janus Benchmark-S (IJB-S), TinyFace, DeepChange, Multi-Scene Multi-Time 2017 (MSMT17), and our novel long-distance datasets.

Distorted image restoration using stacked adversarial network

Generative and Discriminative Learning for Distorted Image Restoration.

Iterative Learning for Distorted Image Restoration

A Double Feature Fusion Network with Progressive Learning for Sharper Inpainting

Improving Visual Quality and Transferability of Adversarial Attacks on Face Recognition Simultaneously with Adversarial Restoration

Facial Image Restoration Based on Adversarial Training and Convolutional Neural Network

Stacking Networks Dynamically for Image Restoration Based on the Plug-and-Play Framework

Attention-based Adaptive Selection of Operations for Image Restoration in the Presence of Unknown Combined Distortions

DeepClean: Integrated Distortion Identification and Algorithm Selection for Rectifying Image Corruptions

Blind Geometric Distortion Correction on Images Through Deep Learning

DifFace: Blind Face Restoration with Diffused Error Contraction

DaliID: Distortion-Adaptive Learned Invariance for Identification—A Robust Technique for Face Recognition and Person Re-Identification

ATFaceGAN: Single Face Image Restoration and Recognition from Atmospheric Turbulence

Face2Face: Label-driven Facial Retouching Restoration

Mixed distortion image enhancement method based on joint of deep residuals learning and reinforcement learning

Distorted image restoration via non-rigid registration and lucky-region fusion approach

A Deep Convolutional Encoder–Decoder–Restorer Architecture for Image Deblurring

Dynamic Attentive Graph Learning for Image Restoration

Reconstruction Distortion of Learned Image Compression with Imperceptible Perturbations

Distortion Robust Image Classification using Deep Convolutional Neural Network with Discrete Cosine Transform

RobustFace: a novel image restoration technique for face adversarial robustness improvement