Abstract: Image-based methods have replicated the success from 2-D domain to 3-D point cloud semantic segmentation. However, when we directly apply 2-D techniques to the projected pseudo-image, ...
Abstract: Point scene instance mesh reconstruction is a challenging task since it requires both scene-level instance segmentation and instance-level mesh reconstruction from partial observations ...
This repository contains PyTorch implementation for PoinTr: Diverse Point Cloud Completion with Geometry-Aware Transformers (ICCV 2021 Oral Presentation). PoinTr is a transformer-based model for point ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback