Box-based Refinement for Weakly Supervised and Unsupervised Localization Tasks

Eyal Gomel, Tal Shaharabany, Lior Wolf
2/14/2026

No official code found. Best implementation: eyalgomel/box-based-refinement (6 stars, last commit 2023-09-08).

Abstract

It has been established that training a box-based detector network can enhance the localization performance of weakly supervised and unsupervised methods. Moreover, we extend this understanding by demonstrating that these detectors can be utilized to improve the original network, paving the way for further advancements. To accomplish this, we train the detectors on top of the network output instead of the image data and apply suitable loss backpropagation. Our findings reveal a significant improvement in phrase grounding for the "what is where by looking" task, as well as various methods of unsupervised object discovery. Our code is available at https://github.com/eyalgomel/box-based-refinement.

View on arXivView PDF

Code Implementations(1)

eyalgomel/box-based-refinementCommunity100%

Box-based Refinement for Weakly Supervised and Unsupervised Localization Tasks

60C++, CudaJul 17, 20233 years ago

Similar Papers

MOST: Multiple Object localization with Self-supervised Transformers for object discovery

85% match2026

Cascade-DETR: Delving into High-Quality Universal Object Detection

83% match1 repo1002026

Rethinking 3D Dense Caption and Visual Grounding in A Unified Framework through Prompt-based Localization

82% match2 repos72024

HUWSOD: Holistic Self-training for Unified Weakly Supervised Object Detection

82% match1 repo92024

DETR Does Not Need Multi-Scale or Locality Design

81% match2026

Do Object Detection Localization Errors Affect Human Performance and Trust?

81% match2024

Cite this paper

@article{gomel2026boxbased,
  title  = {Box-based Refinement for Weakly Supervised and Unsupervised Localization Tasks},
  author = {Eyal Gomel and Tal Shaharabany and Lior Wolf},
  year   = {2026},
  eprint = {2309.03874},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/pdf/2309.03874},
  journal = {ICCV 2023 2023}
}

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/2309.03874.svg)](https://codeofpaper.com/paper/2309.03874)

Discussion