I-Corps: An automatic training dataset labelling tool to fill the gap for missing training image datasets

这一i-Corps项目的更广泛的影响/商业潜力是开发一个自动培训数据集标签工具和服务。目前，人工标注费时费力，而且对高质量的训练数据集的需求很高。虽然流行的口语应用程序可以利用大量专门的训练数据集通用标记器团队来帮助创建训练数据集，但对于科学和专业领域来说，生成高质量的训练数据集以馈送到人工智能/机器学习(AI/ML)模型和算法中是一个巨大的挑战。建议的技术使用半自动训练数据集标记工具和在线管理服务，以促进图像数据集的自动裁剪、分割和标记。图像是一种流行的数据类型，广泛应用于气候变化、智慧城市、医疗卫生等领域。提出的技术可以减少数百万数据科学专业人员、研究人员和学生以及公司和其他组织在获取训练数据集方面所花费的时间和资源。该i-Corps项目基于开发软件训练数据集标记工具和在线服务，该工具自动填补缺少高质量训练图像数据集的空白。基于时空人工智能/机器学习(AI/ML)的能力正在开发中，以在一组用户之间自动分类、标记、存储和共享训练数据集。拟议的技术旨在提供一个自动数据标签软件工具，以自动标记数据，并帮助识别和分类数据。该过程可用于创建机器学习模型的训练数据集。具体地说，这是一种可用于对数字图像进行分类和标记的图像标记工具。这项拟议的技术最初是为海冰研究开发的，结果证实，海冰图像可以自动标记。这项技术可能会用于许多领域，包括气候变化和生物医学研究。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of an automatic training dataset labelling tool and service. Currently, labelling is accomplished manually, which is time-consuming, and there is a high demand for high-quality training datasets. While popular spoken language applications can leverage numerous teams of dedicated training dataset general labelers to help create training datasets, it poses a significant challenge for scientific and professional domains to produce high-quality training datasets to feed into Artificial Intelligence/Machine Learning (AI/ML) models and algorithms. The proposed technology uses a semi-automatic training dataset labelling tool and online management service to facilitate automatic cropping, segmenting, and labelling of image datasets. Images are a popular data type that is widely used in climate change, smart cities, medical and health domains. The proposed technology may reduce the time and resources spent on obtaining training datasets by millions of data science professionals, researchers, and students, as well as companies and other organizations.This I-Corps project is based on the development of a software training dataset labelling tool and online service that works automatically to fill the gap of missing high-quality training image datasets. Spatiotemporal Artificial Intelligence/Machine Learning (AI/ML)-based capabilities are under development to automatically classify, label, store, and share training datasets among a group of users. The proposed technology is designed to provide an automated data labeling software tool to automatically tag data and help to identify and classify the data. This process may be used to create training datasets for machine learning models. Specifically, this is an image labelling tool that may be used to classify and label digital images. The proposed technology was developed initially for sea ice research and results confirmed that sea ice images can be automatically labelled. This technology potentially may be used in many areas including climate change and biomedical research.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.