A Cloud-Edge-Device Collaborative Inference Framework

RCS (Rich Communication Suite) is expected to comprehensively upgrade basic short messages to realize the interaction and intelligence of short messages. In this paper, a super-resolution-based thumbnail collaborative inference method is proposed. Aiming at the problems of low image recognition accuracy and prolonged review time on existing platforms, a cloud-edge-device collaborative infrence strategy based on super-resolution is proposed. Deploy the lightweight model on the mobile phone, and use the basic computing power of the terminal to take the lead in reviewing the thumbnails and calculating the confidence; transfer the pictures with low confidence to the side, and use super-resolution to optimize the thumbnails to get better high-definition pictures; finally, the pictures are pushed from the edge to the cloud, and the powerful computing power of the cloud is used for identification using the yolov5 network model to obtain accurate picture identification results and complete the picture collaborative review process. The experimental results show that the framework can effectively improve the efficiency of content review.