SAMCL: Empowering SAM to Continually Learn from Dynamic Domains with Extreme Storage Efficiency

TL;DR

This paper introduces SAMCL, a continual learning method for the Segment Anything Model that efficiently adapts to new domains with minimal forgetting and storage, using modular knowledge decomposition and domain-aware selection.

Contribution

The paper proposes AugModule and Module Selector components that enable efficient, low-storage continual learning for SAM across dynamic domains, addressing catastrophic forgetting and storage challenges.

Findings

SAMCL achieves 0.19% forgetting, outperforming existing methods.

AugModule reduces storage by at least 24.3%.

Module Selector significantly decreases buffer storage requirements.

Abstract

Segment Anything Model (SAM) struggles in open-world scenarios with diverse domains. In such settings, naive fine-tuning with a well-designed learning module is inadequate and often causes catastrophic forgetting issue when learning incrementally. To address this issue, we propose a novel continual learning (CL) method for SAM, termed SAMCL. Rather than relying on a fixed learning module, our method decomposes incremental knowledge into separate modules and trains a selector to choose the appropriate one during inference. However, this intuitive design introduces two key challenges: ensuring effective module learning and selection, and managing storage as tasks accumulate. To tackle these, we introduce two components: AugModule and Module Selector. AugModule reduces the storage of the popular LoRA learning module by sharing parameters across layers while maintaining accuracy. It also employs heatmaps-generated from point prompts-to further enhance domain adaptation with minimal additional cost. Module Selector leverages the observation that SAM's embeddings can effectively distinguish domains, enabling high selection accuracy by training on low-consumed embeddings instead of raw images. Experiments show that SAMCL outperforms state-of-the-art methods, achieving only 0.19% forgetting and at least 2.5% gain on unseen domains. Each AugModule requires just 0.233 MB, reducing storage by at least 24.3% over other fine-tuning approaches. The buffer storage for Module Selector is further reduced by up to 256$\times$.