Advances in Object Manipulation and Simulation

The field of object manipulation and simulation is rapidly advancing, with a focus on developing innovative solutions for complex tasks such as knot detection, 3D object rigging, and garment folding. Researchers are exploring the use of machine learning techniques, such as deep learning and reinforcement learning, to improve the accuracy and efficiency of these tasks. Additionally, there is a growing interest in integrating temporal context and physics-driven approaches to enhance the performance of simulation models. Noteworthy papers in this area include:

  • Anymate: A Dataset and Baselines for Learning 3D Object Rigging, which presents a large-scale dataset and a learning-based auto-rigging framework for 3D objects.
  • WisePanda, a physics-driven deep learning framework for rejoining fragmented ancient bamboo slips, which demonstrates the potential of incorporating physical principles into deep learning models.
  • FoldNet: Learning Generalizable Closed-Loop Policy for Garment Folding, which proposes a synthetic garment dataset and a keypoint-driven asset and demonstration synthesis approach for robotic garment folding.

Sources

Automated Knot Detection and Pairing for Wood Analysis in the Timber Industry

Anymate: A Dataset and Baselines for Learning 3D Object Rigging

Learning Sequential Kinematic Models from Demonstrations for Multi-Jointed Articulated Objects

A Gpu-based solution for large-scale skeletal animation simulation

Hand-Shadow Poser

Beyond Static Perception: Integrating Temporal Context into VLMs for Cloth Folding

Rejoining fragmented ancient bamboo slips with physics-driven deep learning

DLO-Splatting: Tracking Deformable Linear Objects Using 3D Gaussian Splatting

FoldNet: Learning Generalizable Closed-Loop Policy for Garment Folding via Keypoint-Driven Asset and Demonstration Synthesis

Procedural Low-Poly Terrain Generation with Terracing for Computer Games

Built with on top of