EDMB: Edge Detector with Mamba

2025-01-08

AI Summary

This paper addresses two key challenges in edge detection: the difficulty of jointly modeling long-range dependencies and capturing fine-grained edges, and the inefficiency of learning multi-granularity edge representations under single-label supervision. To this end, we propose EDMB, the first edge detector incorporating the state-space model Mamba into edge detection, enabling efficient global–local feature fusion. We further design a fine-grained perception module and a learnable Gaussian distribution decoder to explicitly model multi-granularity edge responses. Additionally, we introduce an evidence lower bound (ELBO)-based loss function to enable effective supervision of multi-granularity predictions using only single-label ground truth. On BSDS500, EDMB achieves 0.837 ODS for single-granularity and 0.851 ODS for multi-granularity evaluation—without multi-scale testing or external data. Cross-domain generalization is validated on NYUDv2 and BIPED.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Transformer-based models have made significant progress in edge detection, but their high computational cost is prohibitive. Recently, vision Mamba have shown excellent ability in efficiently capturing long-range dependencies. Drawing inspiration from this, we propose a novel edge detector with Mamba, termed EDMB, to efficiently generate high-quality multi-granularity edges. In EDMB, Mamba is combined with a global-local architecture, therefore it can focus on both global information and fine-grained cues. The fine-grained cues play a crucial role in edge detection, but are usually ignored by ordinary Mamba. We design a novel decoder to construct learnable Gaussian distributions by fusing global features and fine-grained features. And the multi-grained edges are generated by sampling from the distributions. In order to make multi-granularity edges applicable to single-label data, we introduce Evidence Lower Bound loss to supervise the learning of the distributions. On the multi-label dataset BSDS500, our proposed EDMB achieves competitive single-granularity ODS 0.837 and multi-granularity ODS 0.851 without multi-scale test or extra PASCAL-VOC data. Remarkably, EDMB can be extended to single-label datasets such as NYUDv2 and BIPED. The source code is available at https://github.com/Li-yachuan/EDMB.

Problem

• Efficient Model Design
• Edge and Line Detection
• Image Dataset Diversity

Innovation

• EDMB
• Integrated Feature Learning
• Guided Learning Method