On Label Granularity and Object Localization

Elijah Cole, Kimberly Wilber, Grant Van Horn, Xuan Yang, Marco Fornoni +4 more
7/20/2022
cs.CVcs.LG

No official code found. Best implementation: visipedia/inat_loc (15 stars, MIT, last commit 2023-08-01).

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.

View on arXivView PDF

Code Implementations(1)

visipedia/inat_locCommunity100%
MIT

The iNaturalist Localization Dataset from "On Label Granularity and Object Localization" (ECCV 2022).

152Jul 6, 20223 years agoMIT
computer-visiondatasetdeep-learningeccv2022fine-grained+6 more

Similar Papers

Improving Weakly-Supervised Object Localization Using Adversarial Erasing and Pseudo Label

86% match2024

Generative Prompt Model for Weakly Supervised Object Localization

86% match1 repo572023

TeD-Loc: Text Distillation for Weakly Supervised Object Localization

86% match1 repo12025

Evaluating Weakly Supervised Object Localization Methods Right

85% match1 repo3372020

Task-Specific Context Decoupling for Object Detection

85% match2023

Knowledge-guided Causal Intervention for Weakly-supervised Object Localization

85% match1 repo62023

Cite this paper

@article{cole2022label,
  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},
  year   = {2022},
  eprint = {2207.10225},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2207.10225v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2207.10225.svg)](https://codeofpaper.com/paper/2207.10225)

Discussion