Body Shape Recognition for Online Fashion

我们建议开发一个在线工具，这将给许多人带来潜在的好处，允许未经培训的用户从自己的家中使用标准的数码相机/网络摄像头来实现他们的身体大小和形状。目前，在零售环境中部署的大多数身体扫描技术都很昂贵，需要专门的技术支持，其使用仅限于高端百货公司和专业体育零售商。因此，大多数公众无法享受身体扫描带来的时尚、社会和经济效益。该项目旨在利用低成本的网络摄像头/摄像头将这项技术重新格式化为一项不依赖于专业硬件的服务，并将体验从受控的零售环境重新编程到家庭中。在实现目标之前，我们需要解决技术和文化方面的挑战，我们已经建立了合作伙伴关系来应对这些挑战。我们提出的扫描技术是基于摄影测量，计算机科学的一个分支，旨在从一个或多个2D图像中提取3D信息。这些图像通常是从已知的位置用校准过的相机拍摄的，然而这是我们计划从一开始就从这项技术中去除的两个依赖性。如果已知关于标准体型、结构和姿势的足够的先验知识，包括生理和时尚见解，则可以将工具设计成更智能并且更少依赖于分辨率或多个视图的使用。通过使用创新的计算机视觉技术来构建初始模型的双管齐下的方法，通过应用风格本体和零售商元数据进行细化，我们的目标是创建一个工具，为消费者，零售商和制造商提供利益。我们将与我们的行业合作伙伴密切合作，帮助我们定义数据捕获工具的基本要求。这将包括在线服装品牌，也有生产利益，并正在与Bodymetrics@Selfridges合作，开发定制合身的服装。这种家庭规模工具的潜力，使更广泛的基础设施，快速制造和精简的分销网络，使我们接近一个特定类型的零售商，将大大受益于减少购买回报和更大的市场洞察力。我们计划开发一个原型，用于在我们的行业合作伙伴的网站内进行集成，我们将链接到其库存数据库，我们的初步试验将旨在确定用户群的规模，并从合作伙伴的库存中向他们推荐商品。将根据用户组的反馈来判断结果，并将在获得新知识时重新指定基于网络摄像头的工具。还将对使用网络摄像头产生的用户体验进行定性研究，并将制定鼓励用户的解决方案，并考虑到文化差异以及对信任和安全的看法，作出可能的修改。这项研究为项目团队提供了研究用户体验的可能性，这是扫描技术发展缓慢的重要原因。该项目将有助于促进服装大规模定制的发展，这从根本上说是一种新的商业模式。在这种模式下，服装生产的“按需”为特定客户的适合和风格的喜好。这使零售商能够在产品生产之前从顾客那里获得资金;这是零售业标准现金流动态的“逆转”，为服装零售商建立了一个非常有吸引力的商业模式。该项目还可以通过帮助“降低信息成本”来提高在线零售商目前采用的商业模式的效率。该项目旨在开创的方法可以被视为一种有效的在线“搜索衣服”的方式，因为它有助于将个人的身体与适合该人的衣服相匹配。

We are proposing to develop an online tool that will give potential benefit to many, allowing an untrained user to realise their body size and shape, from their own home, using a standard digital camera/webcam. At present, the majority of body scanning technologies deployed in retail environments are expensive and require dedicated technical support, confining their use to high-end department stores and specialist sports retailers. The fashion, social and economic benefits that body scanning offers are therefore inaccessible to the majority of the general public. This project aims to exploit low cost webcam/cameras to reformat this technology into a service that isn't reliant upon specialist hardware, and to reprogram the experience out from the controlled retail environment and into the home.There are both technical and cultural challenges to be addressed before we can achieve our goal and we have created a collaborative partnership to deal with them. Our proposed scanning technology is based upon photogrammetry, a branch of Computer Science that aims to extract 3D information from one or more 2D images. The images are usually taken from known locations with calibrated cameras, however these are two dependencies that we plan to remove from this technology from the outset. If enough prior knowledge about standard body shape, structure and posture is known, both physiological and fashion insights included, a tool can be designed to be more intelligent and less reliant upon resolution or use of multiple views. Through a double-pronged approach that uses innovative Computer Vision techniques to construct an initial model, to refinement by the application of style ontologies and retailer metadata, we aim to create a tool that will offer benefits to consumer, retailer and manufacturer alike. We will work closely with our industry partner to help us define the base requirements of our data capture tool. This will include online clothing brands that also have production interests and are working with Bodymetrics@Selfridges to develop custom fit clothing. The potential for this home sizing tool to enable a wider infrastructure of rapid manufacturing and leaner distribution networks has led us to approach a particular type of retailer that would benefit greatly from a reduction in purchase returns and greater market insights. We plan to develop a prototype for integration within our industry partner's website, we will link to its stock database and our initial trial will aim to size a user group, and recommend them items from the partner's stock. Results will be judged upon the user group's feedback, and the webcam-based tool will be re-specified as new knowledge is gained. Qualitative research will also be undertaken on the user experience arising from using a webcam and solutions will be developed to encourage the user, with possible amendments made to account for cultural differences and perceptions on trust and security. The research offers the potential for the project team to research the user experience, which is a significant contributor to the slow take up to date of scanning technology. The project will help to catalyse the uptake of mass-customisation of garments, which is fundamentally a new business model. In this paradigm, garments are produced 'on-demand' for a particular customer's fit and style preferences. This allows the retailer to obtain monies from customers before the products are produced; a 'reverse' of the standard cash-flow dynamics of retail and establishing a very attractive business model for clothing retailers.The project can also develop the efficiencies of the current business models practised by on-line retailers by helping to 'lower the cost of information'. The method that the project is aiming to pioneer can be seen as an efficient way to 'search for clothes' on-line as it helps to match an individual's body with garments that fit that person.