Abstract

Non-maximum suppression (NMS) is an essential part of the face detection pipeline based on convolutional neural network (CNN). The common approach for NMS used by face detection is a greedy, locally optimal strategy which is to localize object from a set of candidate locations. However, NMS still has some shortcomings, such as sometimes the detection box has no relationship with high classification score, which leads to misjudge face localization during NMS. In this paper, we observed that NMS implemented on the multi-task convolutional neural network (MTCNN) which is a cascaded convolutional network and enhance NMS based on MTCNN to achieve high performance during face detection and alignment. We employ WIDER FACE as test dataset to evaluate our proposal. The precision and recall curve is adopted with three sub sets at different thresholds. The result shows that our proposal can perform better performance than traditional NMS.

Highlights

  • Face detection and alignment have been widely fundamental applications in computer vision which are to generate bounding boxes for identifying frontal face and assign them classification scores

  • We introduce a new framework represented by multi-task cascaded convolutional networks (MTCNN) to integrate face detection and alignment proposed by Kaipeng Zhang et al(5) The advantages of MTCNN is described as follows: First, MTCNN adopted a new cascade and lightweight convolutional neural network (CNN) framework to combine face detection and alignment for achieving good real time performance

  • Our result shows that our proposal can achieve a high performance based on MTCNN

Read more

Summary

Introduction

Face detection and alignment have been widely fundamental applications in computer vision which are to generate bounding boxes for identifying frontal face and assign them classification scores. Even if Chen et al(4) presented a new state-of- art approach for face detection and alignment. An effective approach which is online hard sample mining was proposed in order to improve the performance. In order to avoid this problem, MTCNN employed greedy non-maximum suppression (NMS) as post-processing stage to obtain final detection boxes. The feature of greedy NMS is to make hard decision and set a fixed threshold to decide the range of suppression. In this case that, face misjudged will be occurred and the average precision of entire network may be dropped.

Related Work
Pitfalls in traditional NMS
Proposal
Environment
Measure metrics
Result analysis
Findings
Conclusions and future work
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call