Research On Optimization Of Deep Learning For Object Detection

Object detection is an important topic in computer vision,which has important application value in the automatic driving robot and other fields.The robustness of the object detection algorithm is usually influenced by the color,shape,posture,deformation and occlusion of the object.In addition,the computational complexity of the object detection algorithm is influenced by the number of objects being detected and the dimensions of the feature vectors that describe the object category.Since 2012,after making breakthrough progress in object classification,deep learning has made remarkable achievements in the field of object detection.However,the high time and space complexity of object detection network based on deep learning has hindered the application in practical products.In order to solve this problem,first of all,the MobileNet classification network is used to optimize the FasterRCNN object detection network.The experimental results on the face detection dataset show that the MobileNet classification network is not suitable for optimizing the Faster-RCNN network.After that,a classification network that combines VGG16 and MobileNet is proposed,and the Faster-RCNN object detection network is optimized using the fusion network.The experimental results on the face detection dataset show that the Faster-RCNN object detection network optimized by the fusion classification network has the advantages of using VGG16 and MobileNet to detect faces in Faster-RCNN object detection network.