Object tracking and segmentation in videos

视频中的对象跟踪和分割

• 批准号: 522300-2018
• 负责人: Wang, Yang
• 金额: $ 1.82万
• 依托单位: University of Manitoba
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=668174
• 关键词: Object; tracking; segmentation; videos

AltumView is a company that provides artificial intelligence and computer vision products and solutions for**businesses and consumers around the world. It focuses on design and manufacturing of machine vision**cameras, surveillance cameras, smart cameras with artificial intelligence and machine learning capabilities.**One area of interest for the company is object tracking and segmentation in videos. Given an input video, the**object of interest is manually annotated (either as a bounding box or a segmentation mask) on the first frame.**The goal is to track and segment this object of interest in every frame in the video. A reliable solution to object**tracking/segmentation can be used in many real-world applications. In this project, my research team will**develop new algorithms for object tracking/segmentation. The company can use the technologies developed in**this project to analyze the large amount of video data that it has access to.**We will study two specific problems in object tracking/segmentation. First, we will develop domain adaptation**techniques for object tracking/segmentation. This will allow us to adapt existing object detectors, so that the**object detector can work well on the specific object in the company's videos. Second, we will develop one-shot**learning approaches for object tracking/segmentation. This will allow us to track/segment an object even if we**do not have existing detectors for this object.**The expected outcomes of this project include new algorithms that can effectively track and segment objects in**any video. The company can use these algorithms to analyze the large amount of video data that it currently**has.

AltumView是一家为全球 ** 企业和消费者提供人工智能和计算机视觉产品和解决方案的公司。它专注于设计和制造机器视觉 ** 摄像机，监控摄像机，具有人工智能和机器学习功能的智能摄像机。该公司感兴趣的一个领域是视频中的对象跟踪和分割。给定一个输入视频，** 感兴趣的对象在第一帧上手动注释（作为边界框或分割掩码）。**目标是在视频的每一帧中跟踪和分割感兴趣的对象。一个可靠的目标跟踪/分割解决方案可以用于许多现实世界的应用。在这个项目中，我的研究团队将开发新的目标跟踪/分割算法。该公司可以使用 ** 本项目开发的技术来分析其可以访问的大量视频数据。**我们将研究对象跟踪/分割中的两个具体问题。首先，我们将开发用于对象跟踪/分割的域自适应 ** 技术。这将使我们能够调整现有的对象检测器，以便 ** 对象检测器可以很好地处理公司视频中的特定对象。其次，我们将开发用于对象跟踪/分割的一次性 ** 学习方法。这将允许我们跟踪/分割对象，即使我们 ** 没有针对该对象的现有检测器。该项目的预期成果包括可以有效跟踪和分割 ** 任何视频中的对象的新算法。该公司可以使用这些算法来分析它目前拥有的大量视频数据。