SBIR Phase I: Scaling Up Open Innovation with Crowd Wisdom and Artificial Intelligence (AI) for Smarter and More Sustainable Fashion

这个小企业创新研究（SBIR）第一阶段项目将开发和利用创新的混合智能，即，大数据和人工智能（AI）技术与大众智慧的独特结合，帮助连接和授权独立设计师和中小型零售商/时尚买家（与供应链合作伙伴一起），帮助将原创，独特，时尚的设计与卓越的服装品质带给时尚消费者。该项目还旨在帮助时装业解决生产过剩和浪费（导致环境问题）方面最困难、最关键和最紧迫的挑战。该项目将通过开发新的深度学习驱动的时尚推荐模型来推进推荐技术和时尚智能，并有效地将人类时尚专家的输入和深度学习预测联合收割机结合起来。这些技术将有助于匹配时尚零售买家和设计师，并考虑到独特性和排他性。 该项目还将帮助评估时装设计的关键方面，如独特性和时尚性，并提供更准确的时装需求和销售预测。关键的技术创新是双重的。首先，一种新型的自我监督和深度学习驱动的时尚推荐引擎将有效地利用异构的时尚数据（图像，文本，行为和销售），以帮助在风格兼容性和其他要求下准确地将时尚买家和制造商与（新）设计相匹配。其次，混合智能引擎将有效地将时尚买家的输入（投票和订单）与深度学习模型进行联合收割机和整合，以帮助衡量时尚的独特性，潮流度和销售预测等，新的设计。该项目可以帮助设计师和零售商跟踪趋势和需求，并保持领先的时尚曲线。这个奖项反映了NSF的法定使命，并已被认为是值得支持的评估使用基金会的知识价值和更广泛的影响审查标准。

This Small Business Innovation Research (SBIR) Phase I project will develop and leverage an innovative hybrid intelligence, i.e., a unique combination of Big Data and artificial intelligence (AI) technologies with the wisdom of crowds, to help connect and empower both independent designers and small-to-medium-sized retailers/fashion buyers (together with supply chain partners), to help bring the original, unique, trendy designs with great garment quality to fashion consumers. The project also aims to help the fashion industry to tackle some of its hardest, most critical, and most urgent challenges in overproduction and waste (resulting in environmental issues). The project will advance recommendation technology and fashion intelligence by developing novel deep learning-powered fashion recommendation models, and effectively combine and integrate human fashion experts’ input and deep learning predictions. These techniques will help match fashion retail buyers and design(er)s, with the consideration of uniqueness and exclusivity. The project will also help evaluate key aspects of the fashion designs, such as uniqueness and trendiness, and provide more accurate predictions on fashion demands and sales. The key technology innovations are two-fold. First, a novel self-supervised and deep learning-powered fashion recommendation engine will effectively utilize the heterogeneous fashion data (images, text, behaviors, and sales) to help accurately match fashion buyers and manufacturers with the (new) design(er)s under style compatibility and other requirements. Second, a hybrid intelligence engine will effectively combine and integrate fashion buyers' input (votes and orders) with deep learning models to help measure fashion uniqueness, trendiness, and sales forecasts, etc., of the new designs. The project can help both designers and retailers track the trends and the demands and stay ahead of the fashion curve.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.