Saved in:
Bibliographic Details
Main Authors: Adamyan, Alen, Čížek, Tomáš, Straka, Matej, Janouskova, Klara, Schmid, Martin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.08548
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915383926063104
author Adamyan, Alen
Čížek, Tomáš
Straka, Matej
Janouskova, Klara
Schmid, Martin
author_facet Adamyan, Alen
Čížek, Tomáš
Straka, Matej
Janouskova, Klara
Schmid, Martin
contents Segment Anything Model 2 (SAM 2) has demonstrated strong performance in object segmentation tasks and has become the state-of-the-art for visual object tracking. The model stores information from previous frames in a memory bank, enabling temporal consistency across video sequences. Recent methods augment SAM 2 with hand-crafted update rules to better handle distractors, occlusions, and object motion. We propose a fundamentally different approach using reinforcement learning for optimizing memory updates in SAM 2 by framing memory control as a sequential decision-making problem. In an overfitting setup with a separate agent per video, our method achieves a relative improvement over SAM 2 that exceeds by more than three times the gains of existing heuristics. These results reveal the untapped potential of the memory bank and highlight reinforcement learning as a powerful alternative to hand-crafted update rules for memory control in visual object tracking.
format Preprint
id arxiv_https___arxiv_org_abs_2507_08548
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SAM2RL: Towards Reinforcement Learning Memory Control in Segment Anything Model 2
Adamyan, Alen
Čížek, Tomáš
Straka, Matej
Janouskova, Klara
Schmid, Martin
Computer Vision and Pattern Recognition
Machine Learning
Segment Anything Model 2 (SAM 2) has demonstrated strong performance in object segmentation tasks and has become the state-of-the-art for visual object tracking. The model stores information from previous frames in a memory bank, enabling temporal consistency across video sequences. Recent methods augment SAM 2 with hand-crafted update rules to better handle distractors, occlusions, and object motion. We propose a fundamentally different approach using reinforcement learning for optimizing memory updates in SAM 2 by framing memory control as a sequential decision-making problem. In an overfitting setup with a separate agent per video, our method achieves a relative improvement over SAM 2 that exceeds by more than three times the gains of existing heuristics. These results reveal the untapped potential of the memory bank and highlight reinforcement learning as a powerful alternative to hand-crafted update rules for memory control in visual object tracking.
title SAM2RL: Towards Reinforcement Learning Memory Control in Segment Anything Model 2
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2507.08548