Download PDFOpen PDF in browser

Parakeet: a C2 Framework for Efficient AI Capability Development

EasyChair Preprint no. 4307

9 pagesDate: October 1, 2020


Command and Control (C2) concepts and practices have always co-evolved with communications and information technology. The recent advances in Artificial Intelligence (AI) have changed the way the information technology world performs and require C2 to quickly adapt in order to take full advantage of AI potential. Deep Learning (DL) has been at the forefront of the recent advances in AI. In the field of computer vision, DL has enabled the development of object detection and classification models that can rival human capabilities in some domains. However, DL algorithms are data hungry. Annotated datasets that are relevant to military operations can be hard to come by. Although open source datasets are easily accessible, they rarely present the following characteristics that are prevalent in military operational context: multiple sensor modalities; degraded images, labelled instances of specialized military equipment, multiple points of view. To fill this gap, there is a need to complement datasets available from open source or allies, with data collection activities. Even though data collection itself requires significant efforts, the data labelling phase is often the most effort-intensive step of dataset creation. It is considered the main bottleneck holding back further AI adoption. Labelling is challenging due to time constraint and to subject matter expert availability for validating dataset quality. In this paper, we propose the Parakeet framework which applies a C2 approach to labelled dataset building that creates the conditions necessary to AI capabilities development and operationalization success. We show that managing machine learning activities with a C2 framework can enable faster object detection and classification models development, which in turn will enable better C2 performance by providing timely situation analysis through the detection, recognition and tracking of objects and activities of interest.

Keyphrases: Artificial Intelligence, data labelling, dataset, Framework, machine learning, Object Detection and Classification

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
  author = {Valerie Lavigne and Mélanie Breton},
  title = {Parakeet: a C2 Framework for Efficient AI Capability Development},
  howpublished = {EasyChair Preprint no. 4307},

  year = {EasyChair, 2020}}
Download PDFOpen PDF in browser