%0 Journal Article
%T 密集人群小尺度人脸目标检测方法研究<br>Research on Small-Scale Face Target Detection Method for Dense Population
%A 赵耀
%A 秦学
%A 王腾
%J Software Engineering and Applications
%P 185-193
%@ 2325-2278
%D 2022
%I Hans Publishing
%R 10.12677/SEA.2022.112020
%X 针对密集人群图像的人脸目标检测，普遍存在检测出的小尺度人脸目标特征少(不足)等问题，本文提出了一种改进YOLO网络的小型预测特征图特征融合的方法。该方法从浅层网络引出特征图，采用改进的DenseNet增强语义特征后，加入到小型预测尺度特征图，用于丰富小型人脸预测尺度的特征语义信息，进而提高小尺度人脸检测效果。在WIDER FACE数据集上对所提方法进行测试，结果表明，所提方法对密集人群小尺度小人脸的检测精度有较好的提升。<br />
For the face target detection of dense crowd images, there are many problems, such as few (insufficient) small-scale face target features. This paper proposes a feature fusion method of small predictive feature map based on improved Yolo network. This method leads out the feature map from the shallow network, uses the improved DenseNet to enhance the semantic features, and adds it to the small-scale prediction scale feature map to enrich the feature semantic information of the small-scale face prediction scale, so as to improve the effect of small-scale face detection. The proposed method is tested on the WIDER FACE dataset. The results show that the proposed method can improve the detection accuracy of small-scale face of dense population.
%K 计算机视觉，人脸检测，YOLO，DenseNet<br>Computer Vision
%K Face Detection
%K YOLO
%K DenseNet
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=49905