Focusing on data imbalance and intraclass variation,an improved pedestrian detection with a cascade of complex peer AdaBoost classifiers is proposed.The series of the AdaBoost classifiers are learned greedily,along wi...Focusing on data imbalance and intraclass variation,an improved pedestrian detection with a cascade of complex peer AdaBoost classifiers is proposed.The series of the AdaBoost classifiers are learned greedily,along with negative example mining.The complexity of classifiers in the cascade is not limited,so more negative examples are used for training.Furthermore,the cascade becomes an ensemble of strong peer classifiers,which treats intraclass variation.To locally train the AdaBoost classifiers with a high detection rate,a refining strategy is used to discard the hardest negative training examples rather than decreasing their thresholds.Using the aggregate channel feature(ACF),the method achieves miss rates of 35%and 14%on the Caltech pedestrian benchmark and Inria pedestrian dataset,respectively,which are lower than that of increasingly complex AdaBoost classifiers,i.e.,44%and 17%,respectively.Using deep features extracted by the region proposal network(RPN),the method achieves a miss rate of 10.06%on the Caltech pedestrian benchmark,which is also lower than 10.53%from the increasingly complex cascade.This study shows that the proposed method can use more negative examples to train the pedestrian detector.It outperforms the existing cascade of increasingly complex classifiers.展开更多
A real-time pedestrian detection and tracking system using a single video camera was developed to monitor pedestrians. This system contained six modules: video flow capture, pre-processing, movement detection, shadow ...A real-time pedestrian detection and tracking system using a single video camera was developed to monitor pedestrians. This system contained six modules: video flow capture, pre-processing, movement detection, shadow removal, tracking, and object classification. The Gaussian mixture model was utilized to extract the moving object from an image sequence segmented by the mean-shift technique in the pre-processing module. Shadow removal was used to alleviate the negative impact of the shadow to the detected objects. A model-free method was adopted to identify pedestrians. The maximum and minimum integration methods were developed to integrate multiple cues into the mean-shift algorithm and the initial tracking iteration with the competent integrated probability distribution map for object tracking. A simple but effective algorithm was proposed to handle full occlusion cases. The system was tested using real traffic videos from different sites. The results of the test confirm that the system is reliable and has an overall accuracy of over 85%.展开更多
针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘...针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘端人车检测的算法加速。YOMANet算法的主干网络采用轻量型网络MobileNetv2以大幅压缩模型参数量,颈部网络使用深度可分离卷积来代替常规卷积以提升训练速度,并在头部网络嵌入基于归一化的注意力模块(NAM)以增强网络对细节信息的捕获能力。为将YOMANet算法部署到现场可编程门阵列(FPGA)平台,该文针对卷积运算在任务层设计循环分块以调整内循环和外循环的顺序,在运算层对处理引擎单元(PE)设计乘加树,使得多个乘加运算可以同时执行,提高数据的并行计算效率。同时,针对数据存储过程采用双缓存机制来减少数据传输时延,对权重参数和激活函数进行int8数据量化以降低资源消耗。实验结果表明,YOMANet算法在训练平台上的检测精度和检测速度表现优异,对小目标和遮挡目标具备较好的检测能力,有效减少了误检和漏检情况的发生。算法部署到硬件平台后,YOMANet-Accel的目标检测效果保持在较高水平,硬件资源的能效比表现良好,有效发挥了FPGA的并行优势。展开更多
针对实时行人检测场景存在遮挡、形态姿势不同的行人目标,YOLOv5模型对于这些目标检测有明显的漏检问题,提出一种像素差异度注意力机制(pixel difference attention,PDA),不同于传统的通道注意力机制用全局均值池化(global average pool...针对实时行人检测场景存在遮挡、形态姿势不同的行人目标,YOLOv5模型对于这些目标检测有明显的漏检问题,提出一种像素差异度注意力机制(pixel difference attention,PDA),不同于传统的通道注意力机制用全局均值池化(global average pooling,GAP)、全局最大值池化(global max pooling,GMP)来概括整张特征图的信息,全局池化将空间压缩成一个值来表征整个通道,造成了空间信息的流失,PDA将空间信息沿高和宽分别压缩,并将其分别与通道信息联系起来做注意力加权操作,同时提出一种新的通道描述指标表征通道信息,增强空间信息与通道信息的交互,使模型更容易关注到综合了空间和通道维度上的特征图的重要信息,在主干网络末端插入PDA后使模型平均精度(mean average precision,mAP)0.5提升了2.4个百分点,mAP0.5:0.95提升了4.4个百分点;针对实时检测场景的部署和检测速度要求模型拥有较少的参数量和计算量,因此提出了新的轻量化特征提取模块AC3代替原YOLOv5模型中的C3模块,该模块使插入PDA后的改进模型在精度仅仅损失0.2个百分点的情况下,参数量(parameters,Param.)减少了20%左右,浮点运算量(giga floating-point operations,GFLOPs)减少了30%左右。实验结果表明,最终的改进模型比YOLOv5s原模型在VOC行人数据集上mAP0.5提升了2.2个百分点,mAP0.5:0.95提升了3.1个百分点,且参数量减少了20%左右,浮点运算量减少了30%左右,在GTX1050上的检测速度(frames per second,FPS)提升了4。展开更多
基金Project(2018AAA0102102)supported by the National Science and Technology Major Project,ChinaProject(2017WK2074)supported by the Planned Science and Technology Project of Hunan Province,China+1 种基金Project(B18059)supported by the National 111 Project,ChinaProject(61702559)supported by the National Natural Science Foundation of China。
文摘Focusing on data imbalance and intraclass variation,an improved pedestrian detection with a cascade of complex peer AdaBoost classifiers is proposed.The series of the AdaBoost classifiers are learned greedily,along with negative example mining.The complexity of classifiers in the cascade is not limited,so more negative examples are used for training.Furthermore,the cascade becomes an ensemble of strong peer classifiers,which treats intraclass variation.To locally train the AdaBoost classifiers with a high detection rate,a refining strategy is used to discard the hardest negative training examples rather than decreasing their thresholds.Using the aggregate channel feature(ACF),the method achieves miss rates of 35%and 14%on the Caltech pedestrian benchmark and Inria pedestrian dataset,respectively,which are lower than that of increasingly complex AdaBoost classifiers,i.e.,44%and 17%,respectively.Using deep features extracted by the region proposal network(RPN),the method achieves a miss rate of 10.06%on the Caltech pedestrian benchmark,which is also lower than 10.53%from the increasingly complex cascade.This study shows that the proposed method can use more negative examples to train the pedestrian detector.It outperforms the existing cascade of increasingly complex classifiers.
基金Project(50778015)supported by the National Natural Science Foundation of ChinaProject(2012CB725403)supported by the Major State Basic Research Development Program of China
文摘A real-time pedestrian detection and tracking system using a single video camera was developed to monitor pedestrians. This system contained six modules: video flow capture, pre-processing, movement detection, shadow removal, tracking, and object classification. The Gaussian mixture model was utilized to extract the moving object from an image sequence segmented by the mean-shift technique in the pre-processing module. Shadow removal was used to alleviate the negative impact of the shadow to the detected objects. A model-free method was adopted to identify pedestrians. The maximum and minimum integration methods were developed to integrate multiple cues into the mean-shift algorithm and the initial tracking iteration with the competent integrated probability distribution map for object tracking. A simple but effective algorithm was proposed to handle full occlusion cases. The system was tested using real traffic videos from different sites. The results of the test confirm that the system is reliable and has an overall accuracy of over 85%.
文摘针对自动驾驶边缘计算场景中行人车辆检测任务面临的模型计算复杂度高、参数量大导致的部署难题,该文提出一种轻量化神经网络模型YOMANet(Yolo Model Adaptation Network),基于异构FPGA平台设计YOMANet加速器(YOMANet-Accel),实现边缘端人车检测的算法加速。YOMANet算法的主干网络采用轻量型网络MobileNetv2以大幅压缩模型参数量,颈部网络使用深度可分离卷积来代替常规卷积以提升训练速度,并在头部网络嵌入基于归一化的注意力模块(NAM)以增强网络对细节信息的捕获能力。为将YOMANet算法部署到现场可编程门阵列(FPGA)平台,该文针对卷积运算在任务层设计循环分块以调整内循环和外循环的顺序,在运算层对处理引擎单元(PE)设计乘加树,使得多个乘加运算可以同时执行,提高数据的并行计算效率。同时,针对数据存储过程采用双缓存机制来减少数据传输时延,对权重参数和激活函数进行int8数据量化以降低资源消耗。实验结果表明,YOMANet算法在训练平台上的检测精度和检测速度表现优异,对小目标和遮挡目标具备较好的检测能力,有效减少了误检和漏检情况的发生。算法部署到硬件平台后,YOMANet-Accel的目标检测效果保持在较高水平,硬件资源的能效比表现良好,有效发挥了FPGA的并行优势。
文摘针对实时行人检测场景存在遮挡、形态姿势不同的行人目标,YOLOv5模型对于这些目标检测有明显的漏检问题,提出一种像素差异度注意力机制(pixel difference attention,PDA),不同于传统的通道注意力机制用全局均值池化(global average pooling,GAP)、全局最大值池化(global max pooling,GMP)来概括整张特征图的信息,全局池化将空间压缩成一个值来表征整个通道,造成了空间信息的流失,PDA将空间信息沿高和宽分别压缩,并将其分别与通道信息联系起来做注意力加权操作,同时提出一种新的通道描述指标表征通道信息,增强空间信息与通道信息的交互,使模型更容易关注到综合了空间和通道维度上的特征图的重要信息,在主干网络末端插入PDA后使模型平均精度(mean average precision,mAP)0.5提升了2.4个百分点,mAP0.5:0.95提升了4.4个百分点;针对实时检测场景的部署和检测速度要求模型拥有较少的参数量和计算量,因此提出了新的轻量化特征提取模块AC3代替原YOLOv5模型中的C3模块,该模块使插入PDA后的改进模型在精度仅仅损失0.2个百分点的情况下,参数量(parameters,Param.)减少了20%左右,浮点运算量(giga floating-point operations,GFLOPs)减少了30%左右。实验结果表明,最终的改进模型比YOLOv5s原模型在VOC行人数据集上mAP0.5提升了2.2个百分点,mAP0.5:0.95提升了3.1个百分点,且参数量减少了20%左右,浮点运算量减少了30%左右,在GTX1050上的检测速度(frames per second,FPS)提升了4。
