As one of the key technologies of intelligent vehicles, traffic sign detection is still a challenging task because of the tiny size of its target object. To address the challenge, we present a novel detection network ...As one of the key technologies of intelligent vehicles, traffic sign detection is still a challenging task because of the tiny size of its target object. To address the challenge, we present a novel detection network improved from yolo-v3 for the tiny traffic sign with high precision in real-time. First, a visual multi-scale attention module(MSAM), a light-weight yet effective module, is devised to fuse the multi-scale feature maps with channel weights and spatial masks. It increases the representation power of the network by emphasizing useful features and suppressing unnecessary ones. Second, we exploit effectively fine-grained features about tiny objects from the shallower layers through modifying backbone Darknet-53 and adding one prediction head to yolo-v3. Finally, a receptive field block is added into the neck of the network to broaden the receptive field. Experiments prove the effectiveness of our network in both quantitative and qualitative aspects. The m AP@0.5 of our network reaches 0.965 and its detection speed is55.56 FPS for 512 × 512 images on the challenging Tsinghua-Tencent 100 k(TT100 k) dataset.展开更多
在嵌入式设备上,由于算力及存储空间的限制,当前的大型高精度目标检测模型的推理速度较低。为此,本文设计了一种轻量化目标检测模型,用于口罩人脸检测。首先,本文设计了一种高激活性鬼影(High Active Ghost,HAG)模块,以轻量的计算代价...在嵌入式设备上,由于算力及存储空间的限制,当前的大型高精度目标检测模型的推理速度较低。为此,本文设计了一种轻量化目标检测模型,用于口罩人脸检测。首先,本文设计了一种高激活性鬼影(High Active Ghost,HAG)模块,以轻量的计算代价减少特征图中的冗余。其次,利用HAG实现高激活性鬼影跨段部分(High Active Ghost Cross Stage Partial,HAG-CSP)连接模块,提升了跨段部分连接网络结构的特征学习能力。再次,利用HAG-CSP对你只需看一次(You Only Look Once,YOLO)模型进行轻量化改造来得到完整的Ghost-YOLO网络,并构造出一个口罩人脸检测器。实验结果表明,本文提出方法在NVIDIA Jetson NX嵌入式设备上,在检测精度优于其他目标检测算法的前提下,对于640×640的图片,实现了24.72 ms每帧的检测速度,并且减少了模型的参数量。展开更多
针对变电站绝缘套管过热红外图像检测精度不高的问题,提出了基于改进YOLO第7版(you only look once version 7,YOLOv7)算法的检测技术。通过引入改良的跨阶段部分网络幽灵版本3(cross stage partial network ghost version 3,C3Ghost)...针对变电站绝缘套管过热红外图像检测精度不高的问题,提出了基于改进YOLO第7版(you only look once version 7,YOLOv7)算法的检测技术。通过引入改良的跨阶段部分网络幽灵版本3(cross stage partial network ghost version 3,C3Ghost)模块替换头部网络中的扩展高效层聚合网络(extended efficient layer aggregation network,E-ELAN)模块,优化了网络结构,增强了算法对小目标的识别能力。此外,整合了轻量级基于归一化的注意力模块(normalization-based attention module,NAM)到主干网络中以提高对红外图像特征的利用效率,并引入幽灵卷积(ghost convolution,GhostConv)模块替换了网络中的所有卷积,显著降低了模型的大小。结果表明,与YOLOv7初始算法相比,改进YOLOv7算法在F1评分和平均精确率均值上分别提高了19.51%和16.57%,算法的参数量减小了16.3 MB,且检测速度达到了41帧/s,充分证明了该算法在变电站实际应用中的有效性。该研究不仅显著提高了变电站绝缘套管过热红外图像检测的准确性,也能为后续相关技术的研究提供参考。展开更多
基金supported by the National Key R&D Program of China(Grant Nos.2018YFB2101100 and 2019YFB2101600)the National Natural Science Foundation of China(Grant No.62176016)+2 种基金the Guizhou Province Science and Technology Project:Research and Demonstration of Science and Technology Big Data Mining Technology Based on Knowledge Graph(Qiankehe[2021]General 382)the Training Program of the Major Research Plan of the National Natural Science Foundation of China(Grant No.92046015)the Beijing Natural Science Foundation Program and Scientific Research Key Program of Beijing Municipal Commission of Education(Grant No.KZ202010025047)。
文摘As one of the key technologies of intelligent vehicles, traffic sign detection is still a challenging task because of the tiny size of its target object. To address the challenge, we present a novel detection network improved from yolo-v3 for the tiny traffic sign with high precision in real-time. First, a visual multi-scale attention module(MSAM), a light-weight yet effective module, is devised to fuse the multi-scale feature maps with channel weights and spatial masks. It increases the representation power of the network by emphasizing useful features and suppressing unnecessary ones. Second, we exploit effectively fine-grained features about tiny objects from the shallower layers through modifying backbone Darknet-53 and adding one prediction head to yolo-v3. Finally, a receptive field block is added into the neck of the network to broaden the receptive field. Experiments prove the effectiveness of our network in both quantitative and qualitative aspects. The m AP@0.5 of our network reaches 0.965 and its detection speed is55.56 FPS for 512 × 512 images on the challenging Tsinghua-Tencent 100 k(TT100 k) dataset.
文摘在嵌入式设备上,由于算力及存储空间的限制,当前的大型高精度目标检测模型的推理速度较低。为此,本文设计了一种轻量化目标检测模型,用于口罩人脸检测。首先,本文设计了一种高激活性鬼影(High Active Ghost,HAG)模块,以轻量的计算代价减少特征图中的冗余。其次,利用HAG实现高激活性鬼影跨段部分(High Active Ghost Cross Stage Partial,HAG-CSP)连接模块,提升了跨段部分连接网络结构的特征学习能力。再次,利用HAG-CSP对你只需看一次(You Only Look Once,YOLO)模型进行轻量化改造来得到完整的Ghost-YOLO网络,并构造出一个口罩人脸检测器。实验结果表明,本文提出方法在NVIDIA Jetson NX嵌入式设备上,在检测精度优于其他目标检测算法的前提下,对于640×640的图片,实现了24.72 ms每帧的检测速度,并且减少了模型的参数量。
文摘针对变电站绝缘套管过热红外图像检测精度不高的问题,提出了基于改进YOLO第7版(you only look once version 7,YOLOv7)算法的检测技术。通过引入改良的跨阶段部分网络幽灵版本3(cross stage partial network ghost version 3,C3Ghost)模块替换头部网络中的扩展高效层聚合网络(extended efficient layer aggregation network,E-ELAN)模块,优化了网络结构,增强了算法对小目标的识别能力。此外,整合了轻量级基于归一化的注意力模块(normalization-based attention module,NAM)到主干网络中以提高对红外图像特征的利用效率,并引入幽灵卷积(ghost convolution,GhostConv)模块替换了网络中的所有卷积,显著降低了模型的大小。结果表明,与YOLOv7初始算法相比,改进YOLOv7算法在F1评分和平均精确率均值上分别提高了19.51%和16.57%,算法的参数量减小了16.3 MB,且检测速度达到了41帧/s,充分证明了该算法在变电站实际应用中的有效性。该研究不仅显著提高了变电站绝缘套管过热红外图像检测的准确性,也能为后续相关技术的研究提供参考。
