COMPUTER VISION AND MACHINE LEARNING FOR HANDWRITTEN DIAGRAM RECOGNITION
Loading...
Authors
Perez, Carlos A.
Subjects
computer vision
machine learning
diagram recognition
Region Based Convolutional Neural Network
R-CNN
You Only Look Once
YOLO
machine learning
diagram recognition
Region Based Convolutional Neural Network
R-CNN
You Only Look Once
YOLO
Advisors
Kolsch, Mathias N.
Barton, Armon C.
Date of Issue
2025-03
Date
Publisher
Monterey, CA; Naval Postgraduate School
Language
Abstract
Because humans are much better at processing and organizing information visually than narratively, early system schematics or processes will start as handwritten diagrams on either a whiteboard or paper medium. These handwritten diagrams allow for much faster prototyping but still require translation into a digital medium for formal modeling and presentation. Offline graph recognition techniques include deep learning object detectors capable of recognizing diagram symbols. This thesis employs two object classifiers, Faster R-CNN and YOLOv5 to identify and classify objects within handwritten diagrams with an emphasis on detecting and distinguishing between arrow types. We introduce new classes of arrows into an existing data set and fine tune the models to increase classification accuracy. The results indicate that YOLOv5 is superior to Faster R-CNN in the detection of arrows with a precision difference of 0.226 percent. This suggests that YOLOv5 should be the architecture of choice for handwritten diagram recognition programs such as Monterey Phoenix.
Type
Thesis
Description
Series/Report No
Department
Organization
Identifiers
NPS Report Number
Sponsors
National Security Agency, Fort Meade, MD 20755
Funder
Format
Citation
Distribution Statement
Distribution Statement A. Approved for public release: Distribution is unlimited.
Rights
This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States.