Keywords: Dataset pruning, Training-free, Similarity network, Community detection, Polyp segmentation
TL;DR: Training-Free Dataset Pruning via Community Detection in Similarity Networks
Abstract: Recent advances in deep learning have been driven by the availability of larger datasets and more complex models; however, this progress comes at the expense of substantial computational and annotation costs. To address these issues, we introduce a new, training-free dataset pruning method, PRIME, targeting polyp segmentation in medical imaging. To this end, PRIME constructs a similarity network among images in the target dataset and then applies community detection to retain a much smaller, yet representative subset of images from the original dataset. Unlike existing methods that require model training for dataset pruning, our PRIME completely avoids model training, thus significantly reducing computational demands. The reduction in the training dataset reduces 56.2% data annotation costs and enables 2.3$\times$ faster training of polyp segmentation models compared to training on the entire annotated dataset, with only a 0.5% drop in the DICE score. Consequently, our PRIME enables efficient training, fine-tuning, and domain adaptation across medical centers, thus offering a cost-effective solution for deep learning in polyp segmentation. Our implementation is available at https://github.com/SLDGroup/PRIME.
Primary Subject Area: Segmentation
Secondary Subject Area: Application: Endoscopy
Paper Type: Both
Registration Requirement: Yes
Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Created a single midl25_NNN.zip file with midl25_NNN.tex, midl25_NNN.bib, all necessary figures and files., Includes \documentclass{midl}, \jmlryear{2025}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package, Did not use the times package., All authors and co-authors are correctly listed with proper spelling and avoid Unicode characters., Author and institution details are de-anonymized where needed. All author names, affiliations, and paper title are correctly spelled and capitalized in the biography section., References must use the .bib file. Did not override the bibliographystyle defined in midl.cls. Did not use \begin{thebibliography} directly to insert references., Tables and figures do not overflow margins; avoid using \scalebox; used \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., Appendices and supplementary material are included in the same PDF after references., Main paper does not exceed 9 pages; acknowledgements, references, and appendix start on page 10 or later.
Submission Number: 137