Codex will combine the latest automated video mark-up technology with crowdsourced tagging by gamified hobbyist communities to deliver a new way to automatically tag video with objects and people within.
Codex will combine the latest automated video mark-up technology with crowdsourced tagging by gamified hobbyist communities to deliver a new way to automatically tag video with objects and people within.
批准号:
700219
负责人:
金额:
$3.18万
依托单位:
依托单位国家:
英国
项目类别:
GRD Proof of Market
财政年份:
2013
资助国家:
英国
项目状态:
已结题
起止时间:
2013 至 --
中文摘要
广播公司正在积累大量的视频素材档案,代表着数千万小时的节目。为了从价值20亿英镑的视频点播(VOD)和价值8亿英镑的在线视频广告市场中获取更多价值,他们正在投资解锁自己的档案。欧洲和美国广播公司的档案开发预算在2500万至3000万英镑之间,此外,2012年还花费了5亿英镑来委托在线内容。利用这些庞大的档案需要数字化和标签化。前者已经商品化,后者则问题重重。今天流行的手工标记技术是昂贵的,容易出错而且不准确。机器标记算法正在改进,但仍然不准确,需要人类的监督。在成本和时间方面,这两种方法都不可行。Codex将把最新的计算机自动标签技术与游戏化的众包手动标签结合起来,提供一种新的、更精确的方式来创建用于访问视频档案的元数据标签。我们研发的目标是发现最好的技术如何与让爱好者免费标记视频的技术相结合,这样人们、地点、产品和观众的社会足迹就可以在视频中被更快、更准确地标记出来,而且比目前的解决方案成本更低。这项技术将被授权给广播公司批量处理视频,并将允许档案被新的B2C商业模式所利用。Ludifi在提供这项研究方面处于独特的地位。我们开发了可以在网络视频上运行的交互式应用程序,为博客作者提供免费的基于浏览器的工具,让他们的观众可以在社交网络上投票、保存和购买产品、玩游戏和与朋友分享他们的互动。我们对热点和基于时间线的标签进行标记是简单、有效的,但需要手动操作,没有自动标记辅助、对象跟踪或运动传感,也没有任何众包。Codex将改进我们的交互式视频平台,为视频标记提供一种新的、非常强大的工具。
英文摘要
Broadcasters are accumulating vast archives of video footage, representing tens of millions ofhours of programming. They are investing in unlocking their archives to harvest more valuefrom the £2 billion video on demand (VOD) and the £800m online video advertising markets.European and US broadcasters’ archive exploitation budgets range between £25-30 millionpa, in addition to e.£500m spent commissioning online content in 2012. Exploiting these vastarchives requires digitisation and tagging. The former is commoditised, the latter highlyproblematic. Today’s prevalent technique, manual tagging, is expensive, error-prone andinaccurate. Machine tagging algorithms are improving but are still inaccurate, requiringhuman oversight. In cost and time terms, neither alone scales.Codex will combine the latest computer auto-tagging technology with gamified crowd-sourcedmanual tagging to deliver a new, more precise way to create the metadata tags used to accessvideo archives. The goal of our R&D is discovering how best-of-breed technology can workwith techniques that get hobbyists to tag video for free so that people, places, products, andviewers’ social footprints can be tagged inside video faster, more accurately and at lower costthan current solutions. The technology will be licensed to broadcasters to batch process videoand should allow archives to exploited by new B2C business models.Ludifi is uniquely positioned to deliver this research. We create interactive apps that run overweb video, providing free browser-based tools to bloggers to allow their viewers to vote, saveand purchase products, play games and share their interactions with friends over socialnetworks. Our mark-up of hotspots and time-line based tags is simple, effective but manual,with no automated tagging assistance, object tracking or movement sensing, nor anycrowdsourcing. Codex will evolve our interactive video platform to deliver a new, immenselypowerful tool for tagging video.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
海外基金