LRANet++: Low-Rank Approximation Network for Accurate and Efficient Text Spotting

2025-11-08

AI Summary

To address the challenge of balancing accuracy and efficiency in end-to-end arbitrary-shape text spotting, this paper proposes an efficient and robust spotting framework. Methodologically: (1) it introduces a data-driven low-rank subspace model for text geometry, coupled with ℓ₁-norm optimization to achieve noise-robust shape recovery; (2) it designs a parameterized text shape representation and a triple-assignment detection head, jointly leveraging deep-sparse, light-sparse, and dense branches to ensure both training stability and fast inference; (3) it incorporates a lightweight recognition branch to reduce overall computational overhead. Evaluated on multiple arbitrary-shape text benchmarks—including CTW1500 and Total-Text—the method achieves state-of-the-art performance, delivering superior accuracy–latency trade-offs, particularly on curved and irregular text instances.

Technology Category

• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

End-to-end text spotting aims to jointly optimize text detection and recognition within a unified framework. Despite significant progress, designing an accurate and efficient end-to-end text spotter for arbitrary-shaped text remains largely unsolved. We identify the primary bottleneck as the lack of a reliable and efficient text detection method. To address this, we propose a novel parameterized text shape method based on low-rank approximation for precise detection and a triple assignment detection head to enable fast inference. Specifically, unlike other shape representation methods that employ data-irrelevant parameterization, our data-driven approach derives a low-rank subspace directly from labeled text boundaries. To ensure this process is robust against the inherent annotation noise in this data, we utilize a specialized recovery method based on an $ell_1$-norm formulation, which accurately reconstructs the text shape with only a few key orthogonal vectors. By exploiting the inherent shape correlation among different text contours, our method achieves consistency and compactness in shape representation. Next, the triple assignment scheme introduces a novel architecture where a deep sparse branch (for stabilized training) is used to guide the learning of an ultra-lightweight sparse branch (for accelerated inference), while a dense branch provides rich parallel supervision. Building upon these advancements, we integrate the enhanced detection module with a lightweight recognition branch to form an end-to-end text spotting framework, termed LRANet++, capable of accurately and efficiently spotting arbitrary-shaped text. Extensive experiments on several challenging benchmarks demonstrate the superiority of LRANet++ compared to state-of-the-art methods. Code will be available at: https://github.com/ychensu/LRANet-PP.git

Problem

• Develops efficient arbitrary-shaped text detection using low-rank approximation
• Addresses annotation noise robustness through specialized shape reconstruction method
• Integrates lightweight recognition with enhanced detection for end-to-end text spotting

Innovation

• Low-rank approximation for precise text shape detection
• Triple assignment head for fast and stable inference
• Lightweight recognition branch integrated for end-to-end spotting