component detection and labeling

This work addresses the challenges of low detection accuracy, poor segment completeness, and complex post-processing in image line segment detection. We propose a hierarchical line segment detection method based on aligned anchor groups. Our approach leverages multi-level saliency anchors and structured, geometrically aligned anchor groups as initial cues, followed by hierarchical candidate pixel extraction, sequential anchor linking, and dynamic segment refinement to enable continuous line generation. Final output requires only lightweight validation and merging. The core contribution is the introduction of a geometrically aligned anchor group mechanism—replacing conventional end-to-end regression or heuristic post-processing—thereby achieving high localization accuracy while significantly improving segment completeness. Extensive experiments on multiple benchmark datasets demonstrate that our method outperforms state-of-the-art approaches in both precision and recall, without requiring sophisticated optimization strategies.

This work addresses the challenge faced by visually impaired users in accessing node-link diagrams commonly distributed as bitmap images, a task for which existing assistive technologies are ill-suited due to their reliance on structured data rather than visual input. The paper presents the first lightweight deep learning approach for semantic segmentation of such diagram images, training a compact model on a large-scale synthetic dataset to achieve pixel-level parsing. The proposed method attains over 93% pixel accuracy on synthetic data and demonstrates strong performance both quantitatively and qualitatively. By enabling precise extraction of diagram semantics directly from rasterized images, this approach establishes a viable foundation for non-visual interaction and effectively bridges a critical gap in accessibility technology for bitmap-based graphical content.

This work proposes a novel alignment-free shape representation for object classification that overcomes the limitations of traditional descriptor-based methods, which rely on object alignment to compute shape similarity and thus suffer from reduced efficiency and generalization. The approach begins by normalizing object contours, decomposing them into approximately convex segments, and ordering these segments by descending length. A multidimensional geometric feature vector—encompassing segment length, number of extremal points, area, base length, and width—is then extracted to construct a structure-aware representation for similarity measurement and classification. By eliminating the need for explicit alignment, the method achieves competitive classification accuracy across multiple datasets while significantly enhancing the robustness and scalability of shape comparison.

To address performance limitations in object detection and semantic segmentation under complex scenarios—including occlusion, small objects, and cross-domain generalization—this paper proposes a novel multimodal detection paradigm synergizing large language models (LLMs). Methodologically, it systematically integrates CNNs, YOLOv5/v8, and DETR architectures into an LLM-augmented inference framework, augmented by scalable data pipelines, model pruning, and quantization, and evaluated via a multi-dimensional metric system based on mAP and mIoU. Key contributions include: (1) bridging the gap between traditional feature engineering and end-to-end deep learning; (2) introducing a dynamic context enhancement mechanism tailored for challenging environments; and (3) achieving state-of-the-art accuracy-efficiency trade-offs on COCO and ADE20K. The fully open-sourced, reproducible framework significantly improves model generalizability and robustness across diverse real-world conditions.

This work addresses the challenge of accurately segmenting handwritten and printed text in document digitization under the computational constraints of edge devices, where existing deep learning approaches incur high computational costs and are difficult to deploy in lightweight settings. To overcome this, the authors propose a novel lightweight segmentation framework that operates without deep neural networks. The method first extracts semantically coherent text regions via sentence-level connected component segmentation, then introduces a region-aware handwriting descriptor (RHD) to effectively capture the variability inherent in handwriting. Finally, a conventional classifier is employed for efficient discrimination between handwritten and printed text. Evaluated on both the newly curated MAD-HPTS dataset and the public PHD-AS benchmark, the approach outperforms state-of-the-art methods, achieving over 8× inference speedup with only a 1.4% drop in accuracy, thereby substantially reducing computational overhead and enabling practical edge deployment.

Existing segmentation evaluation metrics often lack transparency and modularity, making them ill-suited for diverse tasks such as transparent object, specular surface, or lesion segmentation. This work proposes a unified evaluation framework that decomposes metrics into five modular components: prediction representation, target extraction, target matching, score computation, and metric reporting. For the first time, it systematically analyzes the implicit assumptions and design limitations of mainstream binary segmentation metrics through this modular lens. The framework enables task-aware customization of evaluation protocols, reveals evolutionary trajectories among existing metrics, and is accompanied by an open-source toolkit. By offering a principled and interpretable foundation, this approach paves the way for developing more rational and adaptable segmentation evaluation methodologies.

This study addresses the challenge of label overlap with nodes, edges, and other labels in ordered parallel coordinates plots by proposing a two-stage flexible labeling algorithm that balances layout readability with minimal connection distance. The method first pre-selects fixed-position labels based on readability criteria; for overflow labels, it leverages both internal and external plot space to connect them to their corresponding nodes via non-crossing straight lines. An initial placement is computed using a sparse grid-based cost function, followed by refinement through a force-directed model. Notably, this approach introduces the first strategy specifically tailored to the vertical-edge characteristics of ordered parallel coordinates, flexibly accommodating the dual-region constraints inherent in formal concept analysis and requiring only minor adjustments to suit concept lattice labeling. Experimental results demonstrate that the algorithm efficiently produces non-crossing, highly readable label layouts and has been successfully applied to concept lattice visualization.