Design And Implementation Of Image Feature Detection Service Based On Deep Learning Framework

With the rapid development of computer technology and the widespread spread of social media,digital images and words scattered across Internet platforms have exploded.While providing information and convenience to people’s lives,internet culture also shoulders great challenges.More and more bad information is gradually hidden in all corners of the online world,which greatly endangers the country’s information security,social security,and the physical and mental health of people,especially minors.In addition to eliminating the root causes,for the bad information and pictures that have been scattered on the Internet,the state arranges special personnel to manually verify,eliminate or code them,but the workload is too large.With the continuous advancement of deep learning technology,relying on the detection of image features in this technology,the rate and accuracy of identification have been greatly improved.Although there are many companies on the market that currently operate related testing services,due to the excessive number of services operated,the detection services for sensitive pictures are more like accessories.At the same time,due to the instability of the network,bandwidth limitations and many other factors,there may be a large response delay,which will seriously affect the user experience.How to provide faster and more accurate detection services for sensitive pictures,and distribute large-scale services to limited nodes,increase user satisfaction,and achieve load balancing,which has become a problem that needs to be solved at present.In view of the above problems,in order to achieve multi-framework support and high-performance support,it is necessary to study and optimize the particle swarm optimization algorithm,and apply the algorithm to the scheduling strategy of sensitive image detection services to make the service scheduling more suitable for market needs.The whole thesis will first compare and verify the optimized particle swarm algorithm and the standard particle swarm algorithm through the verification function,and then use Cloud Sim to build a virtual simulation environment to experiment on the optimized particle swarm algorithm and the pre-optimized particle swarm algorithm,and the optimized particle swarm algorithm will be more in line with the needs of the project through experimental data analysis.Through tensor Flow and Py Torch two frameworks and web front-end and back-end technology applications,so that the service can be practical operation,the entire interface provides a deep learning training and inference interface for sensitive images to complete the corresponding image detection services,while the platform also integrates the image preprocessing function,to help users with less engineering volume to complete the pre-processing work in advance,improve the overall detection efficiency.