WP3.5.1 - Development of Datasets for Machine Learning - LongOps

Tender description

As part of the LongOps project, it is required to develop decision support tools capable of helping de-risking operations and supporting training in teleoperated decommissioning tasks involving the detection and pose estimation of objects and debris within nuclear facilities. In recent years the use of Machine Learning (ML) techniques such as object detection, object classification, segmentation, semantic information extraction from images and videos, denoising, etc. has yielded important results in terms of visual and other sensory perception. However, this requires significant volumes of relevant, high-quality annotated data to achieve adequate levels of performance. UKAEA wishes to procure the collection, evaluation, processing, and annotation of data gathered from nuclear-related sites and/or similar scenarios. Lot 1: As part of the LongOps project, it is required to develop decision support tools capable of helping de-risking operations and supporting training in teleoperated decommissioning tasks involving the detection and pose estimation of objects and debris within nuclear facilities. In recent years the use of Machine Learning (ML) techniques such as object detection, object classification, segmentation, semantic information extraction from images and videos, denoising, etc. has yielded important results in terms of visual and other sensory perception. However, this requires significant volumes of relevant, high-quality annotated data to achieve adequate levels of performance. UKAEA wishes to procure the collection, evaluation, processing, and annotation of data gathered from nuclear-related sites and/or similar scenarios. Such data consists primarily of 2D images and 3D point clouds, although other types of data such as temperature, radiation, pressure, etc. may also be considered. The annotated data will then be used to, but not limited to, train machine learning models to perform tasks such as, object detection and classification, semantic and instance segmentation, etc. In addition, this data will be used to develop, evaluate, and validate change detection and anomaly estimation algorithms.