RetinaFace: Single-stage Dense Face Localisation in the Wild

Abstract

Though tremendous strides have been made in uncontrolled face detection, accurate and efficient face localisation in the wild remains an open challenge. This paper presents a robust single-stage face detector, named RetinaFace, which performs pixel-wise face localisation on various scales of faces by taking advantages of joint extra-supervised and self-supervised multi-task learning. Specifically, We make contributions in the following five aspects: (1) We manually annotate five facial landmarks on the WIDER FACE dataset and observe significant improvement in hard face detection with the assistance of this extra supervision signal. (2) We further add a selfsupervised mesh decoder branch for predicting a pixel-wise 3D shape face information in parallel with the existing supervised branches. (3) On the WIDER FACE hard test set, RetinaFace outperforms the state of the art average precision (AP) by 1.1% (achieving AP equal to 91.4%). (4) On the IJB-C test set, RetinaFace enables state of the art methods (ArcFace) to improve their results in face verification (TAR=89.59% for FAR=1e-6). (5) By employing light-weight backbone networks, RetinaFace can run real-time on a single CPU core for a VGA-resolution image. Extra annotations and code have been made available at: https://github.com/deepinsight/ insightface/tree/master/RetinaFace.

Gendered Differences in Face Recognition Accuracy Explained by Hairstyles, Makeup, and Facial Morphology

“…For license reasons, we cannot disclose the COTS name. The input to both matchers are aligned faces resized to 112x112 (ArcFace, genderbalanced matcher) and 224x224 (COTS) using [39], [40]. For ArcFace, 512-d features are extracted, and matched using cosine similarity.…”

LTT-GAN: Looking Through Turbulence by Inverting GANs

“…In addition to those images, we received the corresponding indoor face images without turbulence for reference. We crop and wrap faces with the pre-trained RetinaFace [12] network. The final dataset contains images from 89 separate individuals each having 3 turbulence degraded images in different poses.…”

SVIP: Sequence VerIfication for Procedures in Videos

“…C is the total number of procedures in the training set. Following the paradigm in face verification [4,16,71], we treat sequence verification as a classification task during training. In the testing phase, the embedding distance between two videos indicates the verification score of this pair.…”