[CVPR 2026] AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects

This video presents our paper, "AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects," accepted to the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026, held June 3-7, 2026. The paper was authored by: MERL interns Danrui Li and Jiahao Zhang; MERL researchers Anoop Cherian, Tim K. Marks, Moitreya Chatterjee, and Suhas Lohit; and MERL consultant Prof. Bernard Egger.

Project Page: https://www.merl.com/research/highlights/assemblybench
Paper: https://arxiv.org/abs/2605.12845

Abstract: Assembling objects from parts requires understanding multimodal instructions, linking them to 3D components, and predicting physically plausible 6-DoF motions for each assembly step. Existing datasets focus on simplified scenarios, overlooking shape complexities and assembly trajectories in industrial assemblies. We introduce AssemblyBench, a synthetic dataset of 2,789 industrial objects with multimodal instruction manuals, corresponding 3D part models, and part assembly trajectories. We also propose a transformer-based model, AssemblyDyno, which uses the instructional manual and the 3D shape of each part to jointly predict assembly order and part assembly trajectories. AssemblyDyno outperforms prior works in both assembly pose estimation and trajectory feasibility, where the latter is evaluated by our physics-based simulations.

Видео [CVPR 2026] AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects канала Mitsubishi Electric Research Laboratories (MERL)