面向扶梯不安全行为的改进型深度学习检测算法

Improved Deep Learning Detection Algorithm for Unsafe Escalator Behavior

以YOLOv5s网络模型为基础,引入注意力机制CBAM模块,基于Ghost卷积模块重构网络模型的卷积操作,提出一种面向扶梯不安全行为的改进型深度学习检测算法.然后,在自主收集的扶梯不安全行为数据集上对其进行训练评估.结果表明,所提算法在检测精度有所提高的同时,大幅减少了检测所需的参数量和计算量.

An improved deep learning detection algorithm for unsafe escalator behavior was proposed.The algorithm is based on the YOLOv5s network model,introduces the attention mechanism CBAM module,and reconstructs the convolution operation of the network model based on the Ghost convolution module.It is trained and evaluated on the self-collected escalator unsafe behavior data set.The results show that the proposed algorithm has improved the detection accuracy while greatly reducing the amount of parameters and calculation required for detection.