Research And Application Of Object Detection Algorithm Based On Attention Mechanism

In the field of machine vision,object detection has always been the focus of research.Object detection has wider applications like Unmanned Aerial Vehicle(UAV),Intelligent Drive(ID),and Wise Medical(WM).How to make the algorithm quicker and more accurate is an important problem.In 2012,Alex Net won the ILSVRC classification competition with great advantages,In recent years,deep learning has gradually learned the mainstream research methods and guided the development of artificial intelligence.The object detection algorithm based on convolutional neural network can be divided into two-stage object detection algorithm based on region proposal and single-stage object detection algorithm based on regression prediction.Both algorithms have their own advantages in accuracy or speed.SSD(Single Shot multi-boxes Detection)is one of the best methods with the higher accuracy and quicker speed.Based on it,we introduces attention mechanism as a weak region proposal for enhancing the detected features.(1)This paper introduces the attention mechanism and combines the SSD detection scheme to propose an improved candidate region algorithm(R-SSD)for accuracy.(2)This paper proposes a fast detection algorithm based on multi-scale attention features combined with SSD.On the one hand,the attention mechanism and context information are combined to enhance the detected features;on the other hand,by sharing the multi-scale down-sampling layer,the computational complexity is reduced.