FA-RPN: Floating Region Proposals for Face Detection

TL;DR

This paper introduces FA-RPN, a novel face detection approach that uses a pooling-based region proposal method with an efficient anchor placement strategy, achieving high accuracy on the WIDER dataset.

Contribution

The paper presents FA-RPN, a new face detection framework that improves region proposal quality and computational efficiency through innovative anchor placement and proposal refinement techniques.

Findings

FA-RPN outperforms RPN in face detection proposals.

Achieves 89.4% mAP on WIDER dataset with ResNet-50.

Proposes flexible proposal refinement without retraining.

Abstract

We propose a novel approach for generating region proposals for performing face-detection. Instead of classifying anchor boxes using features from a pixel in the convolutional feature map, we adopt a pooling-based approach for generating region proposals. However, pooling hundreds of thousands of anchors which are evaluated for generating proposals becomes a computational bottleneck during inference. To this end, an efficient anchor placement strategy for reducing the number of anchor-boxes is proposed. We then show that proposals generated by our network (Floating Anchor Region Proposal Network, FA-RPN) are better than RPN for generating region proposals for face detection. We discuss several beneficial features of FA-RPN proposals like iterative refinement, placement of fractional anchors and changing anchors which can be enabled without making any changes to the trained model. Our face detector based on FA-RPN obtains 89.4% mAP with a ResNet-50 backbone on the WIDER dataset.