Search

In this paper, we introduce a novel rate control scheme specifically tailored for live broadcasting scenarios. Notably, in high-definition live transmissions of sports events and video game competitions that typically exceed 1080 p resolution and run at frame rates of 60 fps or higher, the transcoding speed of encoders often becomes a limiting factor, leading to streams with substantial bitrates but unsatisfactory quality metrics. To enhance the overall Quality of Service (QoS) without increasing the bitrate, it is essential to improve the quality of Regions of Interest (ROI).Our proposed solution presents an ROI-based rate reservoir model that ingeniously leverages Convolutional Neural Networks (CNNs) to predict rate control parameters. This approach aims to optimize the bitrate allocation within high bitrate live broadcasts, thus enhancing the image quality within ROIs. Experimental outcomes demonstrate that this algorithm manages to increase the bitrate by no more than 5%, effectively redistributing the reduced bitrate across the entire Group of Pictures (GOP). As a result, it ensures a gradual decrease in the quality of Regions of Uninterest (ROU), thereby maintaining a balanced quality experience throughout the broadcasted content. [ABSTRACT FROM AUTHOR]

With the continuous enhancement of privacy protection globally, there is a problem for the traditional machine learning paradigm, which is that training data cannot be obtained from a single place. Federated learning is considered a viable technique for preserving privacy that can train deep models with decentralized data. Aiming at two-party vertical federated learning, and at common attack problems such as model inversion, gradient leakage, and data theft, we provide a formal definition of Intel SGX's trusted computing base, remote attestation, integrity verification, and encrypted storage, and propose a general federated learning privacy enhancement algorithm in the scenario of a malicious adversary model, and we extend this method to support horizontal federated learning, secure outsourced computation, etc. Furthermore, the method is developed in a Fedlearner framework of open-sourced machine learning to achieve privacy protection of the training data and model without any modification to the existing neural network and algorithm running on the framework. The experimental results show that this scheme substantially improves on the existing schemes in terms of training efficiency, without losing model accuracy. [ABSTRACT FROM AUTHOR]