DIP: Unsupervised Dense In-Context Post-training of Visual Representations

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Sirko-Galouchenko, Sophia, Gidaris, Spyros, Vobecky, Antonin, Bursuc, Andrei, Thome, Nicolas
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908526352269312
author Sirko-Galouchenko, Sophia
Gidaris, Spyros
Vobecky, Antonin
Bursuc, Andrei
Thome, Nicolas
author_facet Sirko-Galouchenko, Sophia
Gidaris, Spyros
Vobecky, Antonin
Bursuc, Andrei
Thome, Nicolas
contents We introduce DIP, a novel unsupervised post-training method designed to enhance dense image representations in large-scale pretrained vision encoders for in-context scene understanding. Unlike prior approaches that rely on complex self-distillation architectures, our method trains the vision encoder using pseudo-tasks that explicitly simulate downstream in-context scenarios, inspired by meta-learning principles. To enable post-training on unlabeled data, we propose an automatic mechanism for generating in-context tasks that combines a pretrained diffusion model and the vision encoder itself. DIP is simple, unsupervised, and computationally efficient, requiring less than 9 hours on a single A100 GPU. By learning dense representations through pseudo in-context tasks, it achieves strong performance across a wide variety of downstream real-world in-context scene understanding tasks. It outperforms both the initial vision encoder and prior methods, offering a practical and effective solution for improving dense representations. Code available here: https://github.com/sirkosophia/DIP
format Preprint
id arxiv_https___arxiv_org_abs_2506_18463
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DIP: Unsupervised Dense In-Context Post-training of Visual Representations
Sirko-Galouchenko, Sophia
Gidaris, Spyros
Vobecky, Antonin
Bursuc, Andrei
Thome, Nicolas
Computer Vision and Pattern Recognition
We introduce DIP, a novel unsupervised post-training method designed to enhance dense image representations in large-scale pretrained vision encoders for in-context scene understanding. Unlike prior approaches that rely on complex self-distillation architectures, our method trains the vision encoder using pseudo-tasks that explicitly simulate downstream in-context scenarios, inspired by meta-learning principles. To enable post-training on unlabeled data, we propose an automatic mechanism for generating in-context tasks that combines a pretrained diffusion model and the vision encoder itself. DIP is simple, unsupervised, and computationally efficient, requiring less than 9 hours on a single A100 GPU. By learning dense representations through pseudo in-context tasks, it achieves strong performance across a wide variety of downstream real-world in-context scene understanding tasks. It outperforms both the initial vision encoder and prior methods, offering a practical and effective solution for improving dense representations. Code available here: https://github.com/sirkosophia/DIP
title DIP: Unsupervised Dense In-Context Post-training of Visual Representations
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.18463