Successful object detection, using CNN, requires lots of well-anno-tated training data which is currently not available for action recognition in hand-ball domain.Augmenting real-world image dataset with synthesized images is not a novel ap-proach, but the effectiveness of the creation of such a dataset and the quantities of generated images required to improve the detection can be.Starting with relatively small training dataset, by combining traditional 3D mod-eling with proceduralism and optimizing generator-annotator pipeline to keep rendering and annotating time under 3 FPS, we achieved 3x better detection re-sults, using YOLO, while only tripling the training dataset.
© Copyright 2019 ICT Innovations 2019. Veröffentlicht von Association for Information and Communication Technologies. Alle Rechte vorbehalten.