YOLO-FaceV2: A Scale and Occlusion Aware Face Detector

TL;DR

YOLO-FaceV2 is a real-time face detection model based on YOLOv5 that improves detection of small, occluded, and tiny faces through novel modules and loss functions, outperforming previous YOLO variants on the WiderFace dataset.

Contribution

The paper introduces RFE, SEAM, and Slide modules, along with NWD Loss, to enhance small face detection, occlusion handling, and class imbalance in a single-stage face detector.

Findings

Outperforms YOLO variants on WiderFace dataset across all difficulty levels.

Effectively detects small and occluded faces in real-time.

Achieves high accuracy with efficient modules and loss functions.

Abstract

In recent years, face detection algorithms based on deep learning have made great progress. These algorithms can be generally divided into two categories, i.e. two-stage detector like Faster R-CNN and one-stage detector like YOLO. Because of the better balance between accuracy and speed, one-stage detectors have been widely used in many applications. In this paper, we propose a real-time face detector based on the one-stage detector YOLOv5, named YOLO-FaceV2. We design a Receptive Field Enhancement module called RFE to enhance receptive field of small face, and use NWD Loss to make up for the sensitivity of IoU to the location deviation of tiny objects. For face occlusion, we present an attention module named SEAM and introduce Repulsion Loss to solve it. Moreover, we use a weight function Slide to solve the imbalance between easy and hard samples and use the information of the effective receptive field to design the anchor. The experimental results on WiderFace dataset show that our face detector outperforms YOLO and its variants can be find in all easy, medium and hard subsets. Source code in https://github.com/Krasjet-Yu/YOLO-FaceV2