On Label Granularity and Object Localization

Elijah Cole; Kimberly Wilber; Grant Van Horn; Xuan Yang; Marco Fornoni; Pietro Perona; Serge Belongie; Andrew Howard; Oisin Mac Aodha

On Label Granularity and Object Localization

Computer Vision and Pattern Recognition 2022-07-22 v1 Machine Learning

Authors: Elijah Cole , Kimberly Wilber , Grant Van Horn , Xuan Yang , Marco Fornoni , Pietro Perona , Serge Belongie , Andrew Howard , Oisin Mac Aodha

View on arXiv ↗ PDF ↗

Abstract

Weakly supervised object localization (WSOL) aims to learn representations that encode object location using only image-level category labels. However, many objects can be labeled at different levels of granularity. Is it an animal, a bird, or a great horned owl? Which image-level labels should we use? In this paper we study the role of label granularity in WSOL. To facilitate this investigation we introduce iNatLoc500, a new large-scale fine-grained benchmark dataset for WSOL. Surprisingly, we find that choosing the right training label granularity provides a much larger performance boost than choosing the best WSOL algorithm. We also show that changing the label granularity can significantly improve data efficiency.

Keywords

weakly supervised semantic segmentation object detection image classification

Cite

@article{arxiv.2207.10225,
  title  = {On Label Granularity and Object Localization},
  author = {Elijah Cole and Kimberly Wilber and Grant Van Horn and Xuan Yang and Marco Fornoni and Pietro Perona and Serge Belongie and Andrew Howard and Oisin Mac Aodha},
  journal= {arXiv preprint arXiv:2207.10225},
  year   = {2022}
}

Comments

ECCV 2022

On Label Granularity and Object Localization

Abstract

Keywords

Cite

Comments

Related papers