| 张海超,张 闯.融合注意力的轻量级行为识别网络研究[J].电子测量与仪器学报,2022,36(5):173-179 |
| 融合注意力的轻量级行为识别网络研究 |
| Research on lightweight action recognition network integrating attention |
| |
| DOI： |
| 中文关键词: 3D 卷积神经网络 行为识别 注意力机制 轻量化 |
| 英文关键词:3D convolutional neural network action recognition attention mechanism light weight |
| 基金项目:国家自然科学基金(61906097)、江苏省高校优势学科项目资助 |
|
|
|
|
| 摘要点击次数: 1088 |
| 全文下载次数: 1304 |
| 中文摘要: |
| 针对传统的三维卷积神经网络存在参数量多、信息冗余和时序信息提取不充分 3 个问题,提出了一种融合注意力的轻
量级行为识别网络。 首先,为轻量化网络参数和融合短中长时序信息,提出了高效残差块来替代两个级联的 3×3×3 卷积;其
次,对通道注意力进行拓展,提出了时间注意力机制,并将两者嵌入在网络中抑制冗余信息对识别结果的影响;最后,在 UCF101
数据集上进行实验验证该网络的有效性。 结果表明,提出的行为识别网络计算成本为 8. 9 GFlops,参数量为 18. 0 M,识别准确
率为 94. 8%,与其他行为识别方法相比,以低成本的计算量实现了较高的识别准确率。 |
| 英文摘要: |
| A lightweight action recognition network with fused attention is proposed to deal with the three problems of the traditional 3D
convolutional neural network: large number of parameters, information redundancy and insufficient extraction of temporal information.
First, in order to lighten the network parameters and fuse short-medium-long temporal information, an efficient residual block is
developed to replace two cascaded 3×3×3 convolutions; second, by extending the channel attention mechanism, a temporal attention
mechanism is derived, and both of the two mechanisms are integrated into the proposed network to suppress the influence of redundant
information on recognition results; finally, experiments are conducted on the UCF101 dataset to verify the effectiveness of the network.
The results show that the proposed action recognition network has a computational cost of 8. 9 GFlops, a parameter amount of 18. 0 M,
and a recognition accuracy rate of 94. 8%, which reveals a high recognition accuracy with a low cost computation in comparison with
other behavior recognition networks. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
