ID 原文 译文
16445 首先将特征提取网络VGG16中Conv4_3层和Conv5_3层提取的特征进行融合形成基础融合特征； Firstly, thefeatures extracted from the Conv4_3 layer and Conv5_3 layer are combined to form basic fusion features.
16446 然后将基础融合特征输入到小型的多尺度语义信息融合模块中，得到具有丰富上下文信息和空间细节信息的语义特征，同时把语义特征和基础融合特征经过特征重利用模块获得浅层增强特征； Then the basic fusion features are inputted into a small multi-scale semantic information fusion module toobtain semantic features of rich contextual information and spatial detail information. The semantic featuresare fused into the basics features by the feature reuse module to obtain shallow enhanced features.
16447 最后基于浅层增强特征进行一系列卷积获取多个不同尺度的特征，并输入各检测分支进行检测，利用非极大值抑制算法实现最终的检测结果。 Finally, aseries of convolutions are performed based on the shallow enhanced features to obtain multiple features withdifferent scales. Multiple detection branches are then constructed based on the features of different scales. The non-maximum suppression algorithm is used to achieve the final detection.
16448 在PASCAL VOC2007和MS COCO2014数据集上进行测试，模型的平均精度均值分别为81.2%和33.7%，相对于经典的单极多盒检测器(SSD)算法，分别提高了2.7%和4.9%； The average accuracy of theproposed model is 81.2% and 33.7% on the PASCAL VOC2007 and MS COCO2014 datasets respectively,which is 2.7% and 4.9% higher than the classic Single Shot multibox Detector (SSD) algorithm.
16449 此外，该文方法在检测小目标和密集目标场景上，检测精度和召回率都有显著提升。 In addition, ondetecting small targets in dense target scenes, the detection accuracy and recall rate of the proposed method aresignificantly improved.
16450 实验结果表明该文算法采用特征金字塔结构增强了浅层特征的语义信息，并利用特征重利用模块有效保留了浅层的细节信息用于检测，增强了模型对小目标和密集目标的检测效果。 The experimental results show that the feature pyramid structure can enhance thesemantic information of shallow features, and the feature reuse module can effectively retain shallow detailinformation for detection, so the proposed method can get better detection performance on small targets anddense targets.
16451 单发多框检测器SSD是一种在简单、快速和准确性之间有着较好平衡的目标检测器算法。 Single Shot multibox Detector (SSD) is a object detection algorithm that provides the optimal trade-off among simplicity, speed and accuracy.
16452 SSD网络结构中检测层单一的利用方式使得特征信息利用不充分，将导致小目标检测不够鲁棒。 The single use of detection layers in SSD network structure makesthe feature information not fully utilized, which will lead to the small object detection are not robust enough.
16453 该文提出一种基于注意力机制的单发多框检测器算法ASSD。 Inthis paper, an Attention based Single Shot multibox Detector (ASSD) is proposed.
16454 ASSD算法首先利用提出的双向特征融合模块进行特征信息融合以获取包含丰富细节和语义信息的特征层，然后利用提出的联合注意力单元进一步挖掘重点特征信息进而指导模型优化。 The ASSD algorithm first uses the proposed two-way feature fusion module to fuse the feature information to obtain the feature layer which containing rich details and semantic information. Then, the proposed joint attention unit is used to minefurther the key feature information to guide the model optimization.