Collaborative Research: ABI Development: A User-friendly Tool for Highly Accurate Video Tracking

生物科学最近取得了很大的进步，不仅利用了自动化的数据收集方法，而且使研究人员能够通过公共数据库有效地分享结果。该项目将通过制作免费软件来高效自动提取运动数据，并将这些数据直接添加到公共数据库（KNB数据库）中，从而实现使用视频跟踪动物，细胞或机器人运动的应用程序。该软件将易于使用，并适应新类型的视频，这些问题迄今为止一直是现有视频跟踪工具广泛采用的障碍。共享运动数据和视频的能力将促进研究人员之间的合作。这两种功能将使动物群体行为、行为遗传学、细胞生物学和集体机器人技术等领域以及其他记录许多个体运动的领域取得根本性的新进展。该软件，由于其易于使用，并能够从数据库中访问研究视频，也将作为一个工具，在教学中的K-12和大学水平。此外，该项目还将培养生物学和计算机科学方面的大学生和研究生，这种跨学科的培训对于当今生物学研究的进步至关重要。该项目将实现清晰的当前图形用户界面设计的独特组合，以提高最先进的机器学习技术的可用性，从而提高运动跟踪的准确性。此外，开发的软件将使用户能够将结果可视化，以供验证和分析，并包括用户纠正任何剩余跟踪错误的功能。这将使用户能够获得科学质量的数据输出，而不必使用多个软件应用程序，也不必手动后处理数据文件。在该项目的范围内，将举办几次讲习班，并开发一个网站，以改善生物学学生和研究人员的可访问性。该项目还将开发与现有KNB科学数据储存库的直接链接，以便用户可以访问储存库，比较其结果或轻松完成元分析。除了推进生物学研究外，这还将为计算机视觉科学家提供广泛的资源，提供大量具有准确用户注释的视频，以改进目标检测和跟踪等核心算法。 更多信息可以在http://www.abctracker.org上找到。

Biological science has made great strides recently both by capitalizing on automated methods of data collection and by enabling researchers to share results efficiently via public databases. This project will achieve the same for applications that use videos to track movement of animals, cells, or robots, by producing freely available software for efficient and automatic extraction of movement data and directly adding such data to a public database (the KNB Data Repository). The software will be easy to use and adapt to new types of videos, issues that have so far been roadblocks to widespread adoption of existing video tracking tools. The ability to share both movement data and videos will stimulate collaboration among researchers. Both functionalities will enable fundamentally new advances in such areas as animal group behavior, behavioral genetics, cell biology, and collective robotics, and other fields that record the movements of many individuals. The software, because of its ease of use and enabled access to research videos from the database, will also serve as a tool in teaching at the K-12 and college level. In addition, this project will serve to train several college and graduate students in both biology and computer science; such interdisciplinary training is essential for advances in biological research today.This project will implement a unique combination of clear, current graphical user interface design to improve usability with state-of-the-art machine learning techniques to improve movement tracking accuracy. In addition, the developed software will enable users to visualize results for validation and analysis, and include functionality for users to correct any remaining tracking errors. This will enable users to get scientific-quality data output without having to employ multiple software applications and without having to manually post-process data files. In the context of the project, several workshops will be held and a website developed to improve accessibility for students and researchers in biology. The project will also develop a direct link to the existing KNB scientific data repository, such that users can access the repository, compare their results, or complete meta-analyses easily. Besides advancing biological research, this will also generate an extensive resource for computer vision scientists by providing a large collection of videos with accurate user annotation for improving core algorithms such as object detection and tracking. More information may be found at http://www.abctracker.org.