Search papers, labs, and topics across Lattice.
The paper introduces AutoSew, a geometry-based approach for predicting stitch correspondences in 2D garment patterns by formulating it as a graph matching problem. They use a Graph Neural Network to encode local and global geometric context of pattern contours and a differentiable optimal transport solver to infer stitching relationships, including multi-edge connections. The method achieves a 96% F1-score and successfully assembles 73.3% of test garments without error, demonstrating the effectiveness of geometry-based stitching prediction.
Ditch manual annotations: AutoSew uses only 2D pattern geometry and GNNs to predict garment stitching with high accuracy, even with complex multi-edge connections.
Automating garment assembly from sewing patterns remains a significant challenge due to the lack of standardized annotation protocols and the frequent absence of semantic cues. Existing methods often rely on panel labels or handcrafted heuristics, which limit their applicability to real-world, non-conforming patterns. We present AutoSew, a fully automatic, geometry-based approach for predicting stitch correspondences directly from 2D pattern contours. AutoSew formulates the problem as a graph matching task, leveraging a Graph Neural Network to capture local and global geometric context, and employing a differentiable optimal transport solver to infer stitching relationships-including multi-edge connections. To support this task, we update the GarmentCodeData dataset modifying over 18k patterns with realistic multi-edge annotations, reflecting industrial assembly scenarios. AutoSew achieves 96% F1-score and successfully assembles 73.3% of test garments without error, outperforming existing methods while relying solely on geometric input. Our results demonstrate that geometry alone can robustly guide stitching prediction, enabling scalable garment assembly without manual input.
