SimSAM: Simple Siamese Representations Based Semantic Affinity Matrix for Unsupervised Image Segmentation

Chanda Grover Kamra, Indra Deep Mastan, Nitin Kumar, Debayan Gupta
6/12/2024
cs.CV

Official code: chandagrover/SimSAM (6 stars, last commit 2024-02-08).

Abstract

Recent developments in self-supervised learning (SSL) have made it possible to learn data representations without the need for annotations. Inspired by the non-contrastive SSL approach (SimSiam), we introduce a novel framework SIMSAM to compute the Semantic Affinity Matrix, which is significant for unsupervised image segmentation. Given an image, SIMSAM first extracts features using pre-trained DINO-ViT, then projects the features to predict the correlations of dense features in a non-contrastive way. We show applications of the Semantic Affinity Matrix in object segmentation and semantic segmentation tasks. Our code is available at https://github.com/chandagrover/SimSAM.

View on arXivView PDF

Code Implementations(1)

chandagrover/SimSAMOfficial100%
60Python, Jupyter NotebookOct 29, 20232 years ago

Similar Papers

CLASP: Adaptive Spectral Clustering for Unsupervised Per-Image Segmentation

84% match2025

Bayesian Self-Training for Semi-Supervised 3D Segmentation

84% match2026

Sparsely Annotated Semantic Segmentation with Adaptive Gaussian Mixtures

84% match1 repo972026

Information-Maximized Soft Variable Discretization for Self-Supervised Image Representation Learning

84% match1 repo32025

Boosting Unsupervised Semantic Segmentation with Principal Mask Proposals

83% match1 repo192024

Unsupervised Universal Image Segmentation

83% match1 repo2332026

Cite this paper

@article{kamra2024simsam,
  title  = {SimSAM: Simple Siamese Representations Based Semantic Affinity Matrix for Unsupervised Image Segmentation},
  author = {Chanda Grover Kamra and Indra Deep Mastan and Nitin Kumar and Debayan Gupta},
  year   = {2024},
  eprint = {2406.07986},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.07986v1}
}

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/2406.07986.svg)](https://codeofpaper.com/paper/2406.07986)

Discussion