讲师
金山云
资深产品专家
李青源,金山云资深产品专家。目前负责视频云创新产品线,从无到有推动“AI+视频”方向:智能高清、画质增强等新产品落地;并针对KSC265产品化重构,实现了头条、虎牙、熊猫等众多头部客户的快速应用。在加入金山云前,先后就职于百度和AI领域创业,擅长平台构建和新方向探索。
演讲: KSC265全链路解决方案及演进
时间
2018-10-20 10:00 - 10:45
地点
主要内容
2018年,短视频行业日益火爆,移动直播行业走向成熟,众多视频公司在美国或香港敲钟上市。如何进一步提高用户播放体验和降低运营成本,成为视频从业公司急需解决的两个核心问题。通过提供全链路、全终端支持H.265编解码的解决方案,帮助用户成本节省50%,内容覆盖超过30%,金山云和业界同仁一起不断推动视频行业迭代更新和发展壮大。
所属专题
视频编解码
编码效率和编码复杂度之间是难以调和的矛盾。新的Codec在不断优化算法来降低编码复杂度,成熟的Codec也在通过Pre-Title等技术来降低比特率。本专题将讨论新的编码策略、算法以及相关应用实践。
同专题的演讲
徐迈
近年来,随着智能终端的发展以及在线视频等新型业务的普及,无线网络中所传输的图像视频数据量呈爆炸式增长的趋势,网络带宽供求矛盾日益尖锐。视频编码理论是突破无线网络带宽瓶颈的重要途径之一。传统编码理论一直沿着数字信号处理的思路演进,难以突破“边际效应”。事实上,当前终端计算能力飞速发展,为解决带宽供求矛盾提供了契机。因此,另一种新的研究思路是从人类视觉感知模型出发,利用机器学习的计算工具,在视频大数据驱动下构建人类视觉感知模型,面向人类用户体验(QoE),研究基于视觉感知模型的视频编码关键技术,以智能媒体协同计算换取无线视频通信带宽。本报告将介绍徐迈课题组在面向QoE的感知视频编码方面的工作,主要包括:(1)基于关注点大数据的视频显著性感知模型;(2)基于感知模型的视频压缩方法;(3)基于深度学习的HEVC视频编码优化;(4)数据驱动下的全景视频质量评价与编码优化。
Faouzi Kossentini
The world is today witnessing a revolutionary transformation in the areas of media delivery, processing and consumption. Not only the internet traffic is already dominated by the exchange of visual information but also the visual cloud has become of the center of such traffic. Towards a fast-growing visual cloud, with underlying mostly-Xeon-populated homogeneous data centers, Intel Corporation is now leading the way with an Open Source SW-centric strategy. Intel is already open sourcing its SVT-HEVC encoder in Q3’2018, and it also plans to open source the SVT-AV1 encoder in Q2’2019. Such will increase adoption and decrease the cost of HEVC and AV1 by our visual cloud customers, helping them to accelerate the growth of their visual cloud applications. We will discuss briefly Intel’s current visual cloud SW strategy as well as provide detailed descriptions of the being-open-sourced SVT-HEVC encoder and the soon-to-be-open-sourced SVT-AV1 encoder. We will also present results that will illustrate the performance-quality tradeoffs of each of the SVT-HEVC and SVT-AV1 encoders. Finally, we will invite our visual cloud customers to participate in the development and growth of the new SVT-HEVC and SVT-AV1 Open Source communities.
李晓波(篱悠)
从13年HEVC正式成为国际标准以来,业界人士都在为HEVC的普及做着不懈的努力。H265为提升压缩的效率,计算复杂度相对于H264而言增加了不少。因此在一些计算资源受限的设备上(如手机端)实现HEVC的实时编解码能力面临着很大的挑战。本次演讲的主旨是向大家分享下手淘在客户端侧实现HEVC实时编解码能力的探索之路。
田栋
Point cloud data has been emerging and popular for various applications including virtual reality (VR), augmented reality (AR), plus many other computer vision problems. Comparing to conventional image/video data, several fundamental challenges need to be addressed, e.g., its large data rate, irregular sampling structure, varying geometric as well as associated attributes. In this talk, we first review recent progress in point cloud compression and an ongoing MPEG standardization attempt. A framework to conduct fast point cloud resampling is then presented based graph signal processing, which could be utilized to scale down the complexity of many point cloud processing tasks. Last, deep neural networks (DNNs) are generalized for native supporting on point clouds as a new type of signal. For one example, a folding-based autoencoder (AE) architecture FoldingNet is introduced with great potentials for point cloud classification tasks. In short, fundamental techniques are highlighted how point clouds could be successful with more efficient compression, processing and understanding.
李大龙
作为承载日播放量几十亿的流媒体移动端平台,腾讯视频持续通过技术体系建设及细节的匠心打磨为每天亿万用户提供优质服务。充当高品质内容资源和高质量用户体验之间的连接器,我们以移动端播放框架的系统梳理为先导,打通前后台的整体升级与提升,逐步建立了「三维论」思想来持续推动我们的工作。此次分享就给大家带来我们在这方面的方法论沉淀与总结,全景式展示腾讯视频的播放体验优化工作。
Debargha Mukherjee
The Alliance for Open Media - a consortium of major Internet companies formed in 2016 with the mission to develop open media formats for the web - closed their first video codec AV1 in June of 2018. AV1 is the best standardized video codec available today that is also royalty-free. This talk will provide a high level overview of the coding tools in AV1, with special attention to the tools and features that are industry-first in standardized codecs. Results on standard tests sets will be provided.