Headline Update
We introduce AssemblyBench, a synthetic dataset of industrial objects with multimodal instruction manuals, corresponding 3D part models, and part assembly trajectories. We also propose a transformer-based model, which uses the instructional data to predict part assembly process.
We form the trajectory prediction problem as a denoising impaint task and design a map-based guidance term for the diffusion process. It generates trajectory predictions that adheres almost perfectly to environmental constraints.
A synthetic data generator for multi-view multi-group multi-person human atomic actions and group activities, which facilitates the learning of human-centered tasks across single-person, multi-person, and multi-group conditions.
