MediaFinder: a content-based recommendation system for user-generated audio-visual content

近年来，深度学习方法的使用已经突破了人工智能的界限，导致许多尖端技术取得了巨大进步，包括自动驾驶汽车、面部识别系统和欺诈检测服务。在Mashtraxx Ltd (MXX)，一个由经验丰富的人工智能研究人员组成的团队将深度学习的力量引入了音乐推荐。这门科学的灵感来自于尖端的计算机视觉技术，该技术成功地解决了诸如探测卫星图像中的物体或识别图像中的个人等复杂任务。MXX开发了一套智能音乐制作、编辑和消费的工具。在我们的新研究中是一个基于内容的音乐推荐系统，它允许用户使用音轨作为查询来搜索大型音乐收藏。通过与社交媒体平台和我们自己的移动应用程序对这项技术的初步讨论，我们已经确定了在线平台对互补视频推荐系统的需求，该系统可以应对以用户生成视频和音乐为中心的更困难的挑战。MediaFinder的提议将是推荐技术的重大变革，它将颠覆多个行业。

In recent years, the use of deep learning methods has pushed the boundaries of AI, leading to immense advances in a number of cutting-edge technologies, including self-driving vehicles, facial recognition systems and fraud detection services. At Mashtraxx Ltd (MXX), a team of experienced AI researchers has brought the power of deep learning to music recommendation. The science is inspired by cutting-edge computer vision technologies, which successfully solved complex tasks such as detecting objects in satellite pictures or recognizing individuals in images.MXX has developed a suite of tools for intelligent music production, editing and consumption. Among our new research is a content-based **music recommendation system,** which allows users to search large music collections using an audio track as a query. Through initial discussions of this technology with social media platforms, and our own mobile app, we have identified an unmet need for a complementary **video recommendation system** for online platforms, which can cope with the more difficult challenges centred around user-generated videos as well as music.This MediaFinder proposal will be a substantial step-change in recommendation technology, which will disrupt several sectors.