English

Joint-Embedding Predictive Architecture for Self-Supervised Learning of Mask Classification Architecture

Computer Vision and Pattern Recognition 2024-07-16 v1

Abstract

In this work, we introduce Mask-JEPA, a self-supervised learning framework tailored for mask classification architectures (MCA), to overcome the traditional constraints associated with training segmentation models. Mask-JEPA combines a Joint Embedding Predictive Architecture with MCA to adeptly capture intricate semantics and precise object boundaries. Our approach addresses two critical challenges in self-supervised learning: 1) extracting comprehensive representations for universal image segmentation from a pixel decoder, and 2) effectively training the transformer decoder. The use of the transformer decoder as a predictor within the JEPA framework allows proficient training in universal image segmentation tasks. Through rigorous evaluations on datasets such as ADE20K, Cityscapes and COCO, Mask-JEPA demonstrates not only competitive results but also exceptional adaptability and robustness across various training scenarios. The architecture-agnostic nature of Mask-JEPA further underscores its versatility, allowing seamless adaptation to various mask classification family.

Keywords

Cite

@article{arxiv.2407.10733,
  title  = {Joint-Embedding Predictive Architecture for Self-Supervised Learning of Mask Classification Architecture},
  author = {Dong-Hee Kim and Sungduk Cho and Hyeonwoo Cho and Chanmin Park and Jinyoung Kim and Won Hwa Kim},
  journal= {arXiv preprint arXiv:2407.10733},
  year   = {2024}
}

Comments

27 pages, 5 figures

R2 v1 2026-06-28T17:41:14.682Z