E-M3RF is an equivariant multimodal framework for assembling fragmented 3D objects from geometric, visual, and textual cues, leveraging SE(3) structure to reduce the search space and achieve state-of-the-art robustness under noise and partial observations.

ReassembleNet is a scalable, multimodal deep learning framework for reconstructing complex fragmented structures, combining learned contour keypoints with diffusion-based pose estimation to achieve substantially more accurate rotation and translation recovery.

This paper proposes the RePAIR dataset that represents a challenging benchmark to test modern computational and data-driven methods for puzzle-solving and reassembly tasks, focusing on complex real-world 2D and 3D application contexts.

A deep learning–based text-line segmentation method that robustly extracts handwritten text lines across diverse scripts, layouts, orientations, and document styles without dataset-specific processing.