摘要：【中文】本发明涉及一种基于分布式的WEB无插件视频直播方法，其包括如下步骤：第一步、创建分布式管理服务，该分布式管理服务与Zookeeper应用程序协调服务配合使用，根据流媒体上报的资源使用情况，采用加权轮询方法实现负载均衡，第二步、创建流媒体服务,获取各种实时直播视频源，根据媒体资源负载情况，创建该流媒体服务，实时接收媒体源推送的媒体资源或者实时向媒体资源拉取媒体资源，第三步、创建H5视频播放器，该H5视频播放器创建两个通道的websocket服务连接，其中，一个用于信令交互形成信令通道，另外一个用于RTP媒体数据交互形成媒体通道，该信令通道用于控制令及心跳管理，该媒体通道用于RTP媒体数据的传输。 【EN】The invention relates to a WEB plug-in-free video live broadcast method based on distribution, which comprises the following steps: the method comprises the steps of firstly, creating a distributed management service, wherein the distributed management service is matched with a Zookeeper application program for use, load balancing is achieved by adopting a weighted polling method according to resource use conditions reported by streaming media, secondly, the streaming media service is created, various real-time live video sources are obtained, the streaming media service is created according to media resource load conditions, media resources pushed by media sources are received in real time or the media resources are pulled to the media resources in real time, thirdly, an H5 video player is created, the H5 video player creates websocket service connection of two channels, one channel is used for signaling interaction to form a signaling channel, the other channel is used for RTP media data interaction to form a media channel, the signaling channel is used for control command and heartbeat management, and the media channel is used for RTP media data transmission.

摘要：【中文】本发明涉及一种大规模人脸识别模型优化系统，其包括应用服务器、人脸算法服务器、比对分析服务器、训练管理服务器、采集接入服务器、多路人脸抓拍摄像机，本发明为克服大规模人脸识别应用场景中的不足，从数据层面，提供一种在应用过程中产生的数据进行多维度分析整理，输出标注数据集，进而对算法进行优化训练与性能评估的系统。 【EN】The invention relates to a large-scale face recognition model optimization system, which comprises an application server, a face algorithm server, a comparison analysis server, a training management server, an acquisition access server and a plurality of paths of face snapshot cameras.