Abstract:Sketch-based 3D model retrieval provides a convenient way for users to search for 3D models by sketches. Traditionally, this task is converted to a sketch-based 2D shape retrieval problem by projecting 3D models to 2D images. Local invariant features have been widely used to tackle this problem. However, it suffers from the lack of global context and easily fails when images of different 3D models share multiple similar regions. In this paper, we propose a joint description by fusing local statistical structures and global spatial features. Our description is invariant to scale, translate and rotation. An improved bag-of-features retrieval framework is applied to explore semantic visual word representations. Besides, a novel relevance feedback scheme which combines weight balancing and query modification is designed to further improve the retrieval performance. We conduct various experiments on the common sketch-based watertight model benchmark. The comparative results show that our approach significantly outperforms three state-of-the-art methods, demonstrating its effectiveness and robustness for sketch-based 3D model retrieval.

Relevance Feedback in Freehand Sketchy Shape Recognition

Pen-based gesture design and recognition for sketch editing

Online composite shape recognition based on relevance feedback

Incremental On-Line Sketchy Shape Recognition with Dynamic Modeling and Relevance Feedback

Sketch-Based Retrieval in Large-Scale Image Database Via Position-Aware Silhouette Matching.

Online Sketch Recognition Using Geometric-Based Classifiers

An on-line sketch recognition algorithm for composite shape

Segmentation of Online Sketching Using Geometric Feature

Sketch-Based 3D Model Retrieval via Multi-feature Fusion

Classified Segmentation of Online Complex Freehand Sketching

On-Line Sketch Recognition Using Direction Feature

Research on Freehand Sketchy Graphics Recognition

Freehand Sketching Interfaces: Early Processing For Sketch Recognition

Enhancing Sketch-Based Image Retrieval by Re-Ranking and Relevance Feedback

Study on Contour-Based Freehand Sketch Retrieval

Freehand Sketch Recognition Based on Temporal Stroke Information

Segmentation of Online Freehand Sketching Based on Speed Feature

Applying Relevance Feedback with a Linear Programming Classifier

Sketch Recognition With Natural Correction And Editing

Sketch Retrieval and Relevance Feedback with Biased SVM Classification.

Relevance feedback for sketch retrieval based on linear programming classification