Non-Maximum Suppression for Unknown Class Objects using Image Similarity

Yoshiaki Homma, Toshiki Kikuchi, Yuko Ozasa

Abstract

As a post-processing step for object detection, non-maximum suppression (NMS) has been widely used for many years. Greedy-NMS, which is one of the most widely used NMS methods, is effective if the class of objects is known but not if the class of objects is unknown. To overcome this drawback, we propose an NMS method using an image similarity index that is independent of learning. Even if the overlap of bounding boxes that locate different objects is large, they are considered to have located different objects if the similarity of the images in the bounding boxes is low. In order to evaluate the proposed method, we built a new dataset containing unknown class objects. Our experimental results show that the proposed method can reduce the rate of undetected unknown class objects when using greedy-NMS.

Download


Paper Citation


in Harvard Style

Homma Y., Kikuchi T. and Ozasa Y. (2021). Non-Maximum Suppression for Unknown Class Objects using Image Similarity.In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP, ISBN 978-989-758-488-6, pages 444-449. DOI: 10.5220/0010240304440449


in Bibtex Style

@conference{visapp21,
author={Yoshiaki Homma and Toshiki Kikuchi and Yuko Ozasa},
title={Non-Maximum Suppression for Unknown Class Objects using Image Similarity},
booktitle={Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP,},
year={2021},
pages={444-449},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010240304440449},
isbn={978-989-758-488-6},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 4: VISAPP,
TI - Non-Maximum Suppression for Unknown Class Objects using Image Similarity
SN - 978-989-758-488-6
AU - Homma Y.
AU - Kikuchi T.
AU - Ozasa Y.
PY - 2021
SP - 444
EP - 449
DO - 10.5220/0010240304440449